Case Study: Making data-driven decisions on the marketing strategy of Bellabeat’s Time watch based on the trends from public fitness tracking data
David Alberto
05-06-2024
Note: This is a case study for the completion of my Google Data Analytics certificate. All data used are real and can be downloaded from Kaggle provided by Mobius. To view the source code to produce this work, please visit my repository on GitHub. This whole case study takes 15-20 minutes to read thoroughly.
Here is the link to the Executive Summary.
Introduction:
Founded in 2014 by Urška Sršen and Sandro Mur, Bellabeat is a high-tech wellness company that manufactures products specifically designed and engineered to improve the health of women. Bellabeat offers various products that can monitor the productivity, activity, and reproductive health of women. While Bellabeat is just a small company, it is successful and is capable of expanding its business in the global smart device market. To do this, Urška Sršen has urged the data analytics team to analyze the fitness data of smart devices so that the company can open up opportunities for growth. In this scenario, I am acting as part of the data analytics team for Bellabeat, tasked to focus on one of Bellabeat’s products as well as gain insights from the analysis I discover. I will ultimately provide my recommendations to the executives to help drive the marketing strategy for the company.
As this is a case study for my Google Data Analytics certificate, I will be incorporating the following phases of the data life cycle: Ask, Prepare, Process, Analyze, Share, and Act.
Phase 1: Ask
1.1 Business Task:
Business Task: Analyze general customer usage of fitness tracking devices to uncover trends and generate insights that can be applied to Bellabeat’s Time watch and improve its overall sales.
1.2 Key Stakeholders:
The following are the key stakeholders:
- Urška Sršen: Bellabeat’s cofounder and Chief Creative Officer
- Sando Mur: Bellabeat’s cofounder, mathematician and a key member of the Bellabeat executive team
- Bellabeat marketing analytics team: A team of data analysts responsible for collecting, analyzing, and reporting
Phase 2: Prepare
2.1 Data Source
Data Source: FitBit Fitness Tracker Data, licensed under CC0 public domain, was downloaded from Kaggle through Mobius. The extracted file contained 18 csv files, 11 of which I used to cross reference data to each other. However, in the end, I will ultimately be using only 8 of the files for my analysis due to duplicate data. Below are the names of the files with a brief description:
| Original File Name | Unique Id’s | Description |
| dailyActivity_merged | 33 | Consolidated version of data from various files (but not all) spanning 31 days |
| dailyCalories_merged | 33 | Calories per day over 31 days |
| dailyIntensities_merged | 33 | Intensities (ex: lightly active, fairly active, etc.) over 31 days |
| dailySteps_merged | 33 | Steps per day over 31 days |
| heartrate_seconds_merged | 14 | Heart rate per second over 31 days |
| hourlyCalories_merged | 33 | Calories burned per hour over |
| hourlyIntensities_merged | 33 | Total intensity and average intensity per hour over 31 days |
| hourlySteps_merged | 33 | Steps per hour over 31 days |
| sleepDay_merged | 24 | Total minutes asleep, total time in bed and how many times user went to bed over 31 days |
| minuteSleep_merged | 24 | Sleep quality (?) per minute and logid (?) over 31 days |
| weightLogInfo_merged | 8 | Daily weight in lbs, kg, and BMI over 31 days |
2.2 Data Privacy and Accessibility
Before moving on to cleaning the data, I must first address key concerns regarding the dataset, such as its privacy and security. Since the data is in the public domain under the CC0 license, I am permitted to use it. It is provided by Mobius and can be downloaded from Kaggle.
2.3 Data Credibility, Integrity and Ambiguity
This data was collected between March 12, 2016 and May 12, 2016 by Amazon Mechanical Turk, a crowdsourcing marketplace that is usually very reliable. However, there were several issues regarding the bias, credibility and integrity of this dataset:
The descriptions claims that thirty eligible FitBit users consented to the submission of personal tracker data.” However, a quick preliminary data check revealed that there were actually thirty-three unique ID’s collected in the data. Was this simply an error in the description with rounding of number? Or did the extra ID’s collected come from individuals who did not consent to the submission of personal tracker data?
The dataset is not only outdated, but it also supposedly contained data only between April 12, 2016 and May 12, 2016. This is in contrast to the description that claimed the data contains data between March 12, 2016 and May 12, 2016.
A sample size of 33 participants is relatively low compared to the billions of women in the world. Other files contained as little as 8 different users with 67 records.This could result in bias and distort the result of the analysis.
It wasn’t stated in the dataset description how the data was collected. Could the participants all have resided in a specific area or country? Were they all from a particular group, such as a fitness class? The sample population may not be representative of the whole female population who would benefit from Bellabeat’s products.
The fields and values are ambiguous; nothing in the description of the dataset clarifies them. The following are questions I would ask from the appropriate stakeholder or source regarding the following files:
-
▸ ‘dailyActivity_merged’: what is the difference between the fields “TotalDistance,’ “TrackerDistance,” and “LoggedActivitiesDistance” Also, are the values measured in kilometers or miles? The same goes for columns ‘SedentaryActiveDistance’ and etc.
-
▸ ‘hourlyIntensities_merged’: what do the values in the columns ‘TotalIntensity’ and ‘AverageIntensity’ represent?
-
▸ ‘minuteSleep_merged’: what do the values in the column ‘Value’ represenent?
-
▸ ‘sleepDay_merged’: The column ‘Time’ only shows strictly ’12:00:00’ and ‘AM.’ Does this mean that the column ‘TotalMinutesAsleep’ doesn’t include the current sleep session from the time “before midnight”? This can provide misleading data since if one were to sleep tonight from 10PM – 6AM and didn’t sleep the next night, the data may only record 6 hours of sleep (from 12AM-6AM) for tomorrow instead of 8 hours of sleep for today and 0 hours tomorrow.
-
In addition, I am only using one dataset to conduct my analysis. Perhaps I could obtain more data elsewhere then aggregate the results but for the purpose of this case study, I will continue my analysis with the data I already have.
2.4 Sample Code to Check Data Credibility and Integrity
Below are samples codes ran in R to identify the FitBit dataset’s credibility and integrity:
2.4.1 Load the data into R Studio
The following code just loads the ‘tidyverse’ package within R to provide various tools that help me conduct my analysis. Note the number of rows (940) for this specific data. This means there are 940 records in total.
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.4.4 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.0
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorssetwd('C:\\Users\\dalea\\Documents\\Coursera\\Google Data Analytics\\FitBit Wellness Tracker Dataset\\Fitabase Data 4.12.16-5.12.16\\No longer needed')
# Create first raw table for pre-processing
FitBit_data_raw <- read_csv('dailyActivity_merged.csv')## Rows: 940 Columns: 15
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): ActivityDate
## dbl (14): Id, TotalSteps, TotalDistance, TrackerDistance, LoggedActivitiesDi...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.2.4.2 Count unique ID’s:
The glimpse() function shows that there are 33 rows. This means that there are 33 unique IDs for this data.
# Create table
daily_activity_unique_id <- unique(select(FitBit_data_raw, Id))
# Check data
glimpse(daily_activity_unique_id)## Rows: 33
## Columns: 1
## $ Id <dbl> 1503960366, 1624580081, 1644430081, 1844505072, 1927972279, 2022484…2.4.3 Check count of column ‘ActivityDate’ (number of records):
The following code checks the amount of records for each particular date. Note the number of rows which represent the different dates as well as the data type ‘date’. The count represents the amount of records for the specific date.
# Group data
date_count <- FitBit_data_raw %>%
group_by(ActivityDate) %>%
summarize(count = n())
# Convert to date format
date_count$ActivityDate <- as.Date(date_count$ActivityDate, format = '%m/%d/%Y')
# Check data to make sure data format is as intended
glimpse(date_count) ## Rows: 31
## Columns: 2
## $ ActivityDate <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-15, 2016-04-…
## $ count <int> 33, 33, 33, 33, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 3…2.4.4 Plot Graph (Number of Counts by Date)
Let’s visualize the amount of records by date into a chart.
# Plot graph
ggplot(date_count, aes(x = ActivityDate, y = count)) +
geom_bar(stat = "identity", fill = "skyblue", color = "black") +
labs(x = "Date", y = "Count", title = "Number of Counts by Date") +
theme_minimal() +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1),
plot.title = element_text(size = 15, hjust = 0.5)) 
As you can see, less and less data were being recorded towards the end of the data collection (2016-05-12). It is best to filter out the last day (2016-05-12). This is because the low records can skew results tremendously and provide misleading data. I think the users just lost motivation to record their progress towards the end of the data collection.
Phase 3: Process
3.1 Data Cleaning Overview
Before I can analyze the dataset, I had to make sure the data is clean. Below are the following actions I took to clean the data:
- Created backup of files
- Transformed data type (Date columns split into date and time columns then transformed data type to ‘date’ and ‘time’)
- Verified transformation to appropriate data types were successful using R
- Removed duplicate data
- Checked for blanks or missing values
- Correct inconsistent formats (none)
- Checked if data needs to be standardized by removing extra/leading/trailing spaces (none)
- Checked for errors and typos
- Documented steps for data cleaning
3.2 Check for Blanks
I opened every .csv file in excel then filtered all the columns and checked for the presence of “Blanks.” There were no blanks in all of the dataset except for the column ‘Fat’ in the file ‘weightLogInfo_merged.’ Since this specific file had only 68 max records and the ‘Fat’ column only had 2 records, I will skip this field from analysis.
3.3 Using BigQuery (SQL) to Clean Large Data
While I can perform data cleaning in R as well, I will showcase my ability to process large files in SQL. In this case, we will clean the file ‘heartrate_seconds_merged’ in BigQuery since it has 2.48 millions of rows of data. Otherwise, using Excel will result in data loss.
After importing the ‘heartrate_seconds_merged.csv’ file into BigQuery, I performed the following steps to clean then download the data:
1) Since using the ‘auto detect’ feature for the schema resulted in job errors, I had to manually enter the schema in JSON format.
2) Created a schema using the following JSON format
3) Used CAST() function to “permanently” transform the data into the correct data type (for some reason, saving the data as .csv then verifying the data types in R showed “string” format for all columns)
4) File size was too large to save the whole data locally (2,483,658 rows)
5) Exported the data as a table in BigQuery to save the whole data
6) I then exported the table to Google Cloud Storage so I could download the whole data
7) Data was split into multiple .csv files so I downloaded all of them for analysis in R
8) Combined each .csv file into one large .csv file using R
9) Double check that the new file ‘heartrate_seconds_merged_cleaned.csv’ contains the original 2,483,658 rows but with the correct format and data types this time
3.4 Cross Referencing Files
3.4.1 To verify that the values between the files ‘dailyActivity_merged’ and ‘dailyIntensities_merged’ are the same
The output of the code below reveals that the data from ‘dailyIntensities_merged’ is mostly the same as a portion of the data from ‘dailyActivity_merged.’
setwd('C:/Users/dalea/Documents/Coursera/Google Data Analytics/FitBit Wellness Tracker Dataset/Fitabase Data 4.12.16-5.12.16/No longer needed')
# Create intensities table
FitBit_intensities <- read_csv('dailyIntensities_merged.csv')## Rows: 940 Columns: 10
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): ActivityDay
## dbl (9): Id, SedentaryMinutes, LightlyActiveMinutes, FairlyActiveMinutes, Ve...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.# Create intensities table from raw table (see 2.4.1)
intensities_from_raw <- select(FitBit_data_raw, Id, ActivityDate, SedentaryMinutes, LightlyActiveMinutes, FairlyActiveMinutes, VeryActiveMinutes, SedentaryActiveDistance, LightActiveDistance, ModeratelyActiveDistance, VeryActiveDistance)
# Create duplicate of intensities table but following specific order of columns
intensities_from_intensities <- select(FitBit_intensities, Id, ActivityDay, SedentaryMinutes, LightlyActiveMinutes, FairlyActiveMinutes, VeryActiveMinutes, SedentaryActiveDistance, LightActiveDistance, ModeratelyActiveDistance, VeryActiveDistance)
# Compare the two previous tables to check for exact match
result_intensities <- unique(intensities_from_raw == intensities_from_intensities)
result_intensities## Id ActivityDate SedentaryMinutes LightlyActiveMinutes
## [1,] TRUE TRUE TRUE TRUE
## [2,] TRUE TRUE TRUE TRUE
## [3,] TRUE TRUE TRUE TRUE
## [4,] TRUE TRUE TRUE TRUE
## [5,] TRUE TRUE TRUE TRUE
## [6,] TRUE TRUE TRUE TRUE
## [7,] TRUE TRUE TRUE TRUE
## [8,] TRUE TRUE TRUE TRUE
## [9,] TRUE TRUE TRUE TRUE
## [10,] TRUE TRUE TRUE TRUE
## [11,] TRUE TRUE TRUE TRUE
## [12,] TRUE TRUE TRUE TRUE
## [13,] TRUE TRUE TRUE TRUE
## [14,] TRUE TRUE TRUE TRUE
## FairlyActiveMinutes VeryActiveMinutes SedentaryActiveDistance
## [1,] TRUE TRUE TRUE
## [2,] TRUE TRUE TRUE
## [3,] TRUE TRUE TRUE
## [4,] TRUE TRUE TRUE
## [5,] TRUE TRUE TRUE
## [6,] TRUE TRUE FALSE
## [7,] TRUE TRUE FALSE
## [8,] TRUE TRUE TRUE
## [9,] TRUE TRUE TRUE
## [10,] TRUE TRUE FALSE
## [11,] TRUE TRUE FALSE
## [12,] TRUE TRUE TRUE
## [13,] TRUE TRUE FALSE
## [14,] TRUE TRUE FALSE
## LightActiveDistance ModeratelyActiveDistance VeryActiveDistance
## [1,] FALSE FALSE FALSE
## [2,] FALSE FALSE TRUE
## [3,] FALSE TRUE FALSE
## [4,] TRUE TRUE TRUE
## [5,] FALSE TRUE TRUE
## [6,] FALSE TRUE TRUE
## [7,] FALSE FALSE FALSE
## [8,] TRUE FALSE TRUE
## [9,] TRUE FALSE FALSE
## [10,] FALSE FALSE TRUE
## [11,] FALSE TRUE FALSE
## [12,] TRUE TRUE FALSE
## [13,] TRUE TRUE FALSE
## [14,] TRUE FALSE FALSELet’s investigate why some fields are not an exact match and are showing up FALSE on the comparison test.
I’ll start with a logical comparison test from the common column ‘SedentaryActiveDistance.’
# Extract columns from both tables and insert to two new tables
activities_from_raw <- select(FitBit_data_raw, SedentaryActiveDistance)
activities_from_intensities <- select(FitBit_intensities, SedentaryActiveDistance)
# Check if FALSE value is present
result_activity <- unique(activities_from_raw == activities_from_intensities) #Compares values
result_activity## SedentaryActiveDistance
## [1,] TRUE
## [2,] FALSETo figure out why the logical value from certain rows result in FALSE, I had to locate which rows resulted in the FALSE logical value. I modified the previous code to the following:
# Convert to data frame and add new column with row numbers to locate problematic rows
result_activity <- data.frame(activities_from_raw == activities_from_intensities) %>%
mutate(row_number = row_number())
head(result_activity)## SedentaryActiveDistance row_number
## 1 TRUE 1
## 2 TRUE 2
## 3 TRUE 3
## 4 TRUE 4
## 5 TRUE 5
## 6 TRUE 6# Filter for fields that return FALSE to determine row numbers
false_only <- result_activity %>%
filter(SedentaryActiveDistance == FALSE)
head(false_only)## SedentaryActiveDistance row_number
## 1 FALSE 34
## 2 FALSE 37
## 3 FALSE 38
## 4 FALSE 41
## 5 FALSE 43
## 6 FALSE 45I located these rows in excel and noticed that the rows that contained FALSE values all contained decimals (Ex: Cell J35). See the image below:
I realized that the issue was likely due to floating-point precision differences between the values stored in the Excel files and how R represents those values internally. To fix this, I can simply use the near() function to provide the level of tolerance for decimals.
# Create tolerance up to 2 digits (least number of decimals)
tolerance <- 1e-2
# Sample calculation
c <- 0.00999999977648258
d <- 0.01
# Output is 'TRUE' because value of c rounds up to 0.01 since tolerance is 2 digits
e <- near(d,c, tolerance)
e## [1] TRUE# Compare the values again with tolerance of 2 digits
check_logical <- near(intensities_from_raw$SedentaryActiveDistance, intensities_from_intensities$SedentaryActiveDistance, tolerance)
check_logical## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [16] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [31] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [46] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [61] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [76] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [91] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [106] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [121] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [136] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [151] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [166] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [181] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [196] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [211] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [226] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [241] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [256] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [271] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [286] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [301] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [316] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [331] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [346] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [361] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [376] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [391] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [406] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [421] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [436] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [451] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [466] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [481] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [496] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [511] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [526] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [541] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [556] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [571] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [586] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [601] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [616] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [631] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [646] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [661] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [676] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [691] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [706] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [721] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [736] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [751] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [766] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [781] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [796] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [811] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [826] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [841] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [856] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [871] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [886] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [901] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [916] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [931] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUEAlthough we’ve determined the cause of the discrepancies, we don’t actually need the data for the distance columns like ‘SedentaryActiveDistance.’ As mentioned in section 2.3, it is unclear what the values in these columns represent. We will not be analyzing these columns for this case study.
3.4.2 To verify that the values between ‘dailyActivity_merged’ and ‘dailyCalories_merged’ are the same
# Create table
FitBit_calories <- read_csv('dailyCalories_merged.csv')
# Create different tables to compare the two
calories_from_raw <- select(FitBit_data_raw, Calories)
calories_from_calories <- select(FitBit_calories, Calories)
# Compare the tables to check for exact match
result_calories <- unique(calories_from_raw == calories_from_calories)
result_caloriesThe result of the code shows that there is only one unique value (‘TRUE’) with the logical comparison test for the calories column. This means the two columns from the two tables are an exact match.
## Rows: 940 Columns: 3
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): ActivityDay
## dbl (2): Id, Calories
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.## Calories
## [1,] TRUE3.4.2 To verify that the values between ‘dailyActivity_merged’ and ‘dailySteps_merged’ are the same
FitBit_steps <- read_csv('dailySteps_merged.csv')
steps_from_raw <- select(FitBit_data_raw, TotalSteps)
steps_from_steps <- select(FitBit_steps, StepTotal)
result_steps <- unique(steps_from_raw == steps_from_steps)
result_stepsThe result of the code shows that there is only one unique value (‘TRUE’) with the logical comparison test for the steps column. This means the two columns from the two tables are an exact match.
## Rows: 940 Columns: 3
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): ActivityDay
## dbl (2): Id, StepTotal
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.## TotalSteps
## [1,] TRUE3.5 Remove Duplicate Entries
I used the ‘Remove Duplicates’ Tool from Excel to remove duplicate entries for the following two files:
- ‘sleepDay_merged’
- ‘minuteSleep_merged’
3.6 Data Transformation and Conversion
3.6.1 Cleaning the file ‘dailyActivity_merged’
I transformed and converted the data type ‘ActivityDate’ column from ‘dailyActivity_merged’ file to date format then created new file ‘dailyActivity_merged_cleaned’
Steps:
saved the original ‘dailyActivity_merged’ .csv file as another .csv file with different name ‘dailyActivity_merged_cleaned.’ This will now be the file that I’ll be using for analysis
split the values in the ActivityDate field using “/” as the delimiter to separate the month, day and year into separate columns using the text-to-column tool
created a new column ‘T’ that combined the values from the new columns in a date format
replaced the old ActivityDate column with the values from column ‘T’ then renamed new column name to ‘Date’
deleted the ‘transition columns’ (columns Q, R, S and T) as they are no longer needed
3.6.2 Cleaning the file ‘sleepDay_merged’
I split the values in column ‘SleepDay’ from ‘sleepDay_merged’ according to day and time then transformed the data type to date and time
Steps:
created a duplicate file then name it to ‘sleepDay_merged_cleaned.’ This is now the file I’ll be using for analysis
split the ‘SleepDay’ column into three new columns ‘Date’, ‘Time’ and ‘AM_or_PM’ using the Text to Columns tool then transformed the values into the correct day/time format
3.6.3 Cleaning the file ‘minuteSleep_merged’
I split the values in ‘date’ column from ‘minuteSleep_merged’ file into three columns with the correct data types then capitalized column names for consistency
Steps:
duplicated file then named it ‘minuteSleep_merged_cleaned.’ This is the file I’ll be using now for analysis
split the values in ‘Date’ column into three columns ‘Date’, ‘Time’, and ‘AM_or_PM’ then changed data type accordingly
modified the column names by capitalizing the first letter to be consistent with other columns from other files
The image below shows the final cleaned version:
3.6.4 Cleaning the file ‘weightLogInfo_merged’
I split the values in ‘Date’ column from ‘weightLogInfo_merged’ file into three columns with the correct data types
Steps:
created duplicate file with filename ‘weightLogInfo_merged_cleaned.’ This will now be the new file I’ll be using to conduct my analysis
split the values in ‘Date’ column into three columns ‘Date’, ‘Time’, and ‘AM_or_PM’ then changed data type accordingly
note: this is the only file that contains blanks or missing values. I will not be analyzing the column ‘Fat’
The image below is the final cleaned version of the file:
3.6.5 Cleaning the files ‘hourlyIntensities_merged’ and ‘hourlyCalories_merged’
I splitting the date values into ‘Date’, ‘Time’, and ‘AM_or_PM’ for ‘hourlyIntensities_merged’ and ‘hourlyCalories_merged’
I repeated the cleaning process of splitting the date values into ‘date’, ‘time’ and ‘AM_or_PM’ columns then transformed the data type accordingly for the following files:
- hourlyIntensities_merged
- hourlyCalories_merged
Steps:
I created a duplicate file then renamed it as the originally file but adding ‘_cleaned’ to the end of it
performed the same data transformation steps as in the previous examples then verified in R that the columns are of the correct data type
Phase 4: Analyze & Share
4.1 Overview of Files to Analyze
Now that I’ve done all the necessary cleaning for all the data, I’m left with the following eights files to work with for the analysis:
- dailyActivity_merged_cleaned
- hourlyCalories_merged_cleaned
- hourlyIntensities_merged_cleaned
- hourlySteps_merged_cleaned
- minuteSleep_merged_cleaned
- sleepDay_merged_cleaned
- weightLogInfo_merged_cleaned
- heartrate_seconds_merged_cleaned
4.2 Load all the cleaned files and R packages
Here, you can check the metadata of all the cleaned files I will be analyzing, from the number of rows to the data types of each column.
I’ve already installed the packages ‘tidyverse’ and ‘knitr’ so I don’t need to include that code below. If you are replicating the steps I took, you will need to install the appropriate packages using the install.packages() function. Example: install.packages(‘tidyverse’).
library(tidyverse)
library(knitr)
setwd('C:\\Users\\dalea\\Documents\\Coursera\\Google Data Analytics\\FitBit Wellness Tracker Dataset\\Fitabase Data 4.12.16-5.12.16')
daily_activity <- read_csv('dailyActivity_merged_cleaned.csv')## Rows: 940 Columns: 15
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## dbl (14): Id, TotalSteps, TotalDistance, TrackerDistance, LoggedActivitiesD...
## date (1): Date
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.heart_rate <- read_csv('heartrate_seconds_merged_cleaned.csv')## Rows: 2483658 Columns: 5
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (2): Id, Value
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.hourly_calories <- read_csv('hourlyCalories_merged_cleaned.csv')## Rows: 22099 Columns: 5
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (2): Id, Calories
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.hourly_intensities <- read_csv('hourlyIntensities_merged_cleaned.csv')## Rows: 22099 Columns: 6
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (3): Id, TotalIntensity, AverageIntensity
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.hourly_steps <- read_csv('hourlySteps_merged_cleaned.csv')## Rows: 22099 Columns: 5
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (2): Id, StepTotal
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.minutes_sleep <- read_csv('minuteSleep_merged_cleaned.csv')## Rows: 187978 Columns: 6
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (3): Id, Value, LogId
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.daily_sleep <- read_csv('sleepDay_merged_cleaned.csv')## Rows: 410 Columns: 7
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (4): Id, TotalSleepRecords, TotalMinutesAsleep, TotalTimeInBed
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.weight_log <- read_csv('weightLogInfo_merged_cleaned.csv')## Rows: 67 Columns: 10
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): AM_or_PM
## dbl (6): Id, WeightKg, WeightPounds, Fat, BMI, LogId
## lgl (1): IsManualReport
## date (1): Date
## time (1): Time
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.4.3 Analyzing ‘dailyActivity_merged_cleaned’
4.3.1 FitBit User Types on Daily Usage
Before analyzing our data, it would be helpful to know what proportion of users consistently track their activity using FitBIt both throughout the day and month.
The table I created below determines what proportion of users use the FitBit device for specific durations throughout the day. The table is grouped by ‘Id.’
# Create table on average daily usage of FitBit grouped by 'Id'
daily_usage <- daily_activity %>%
group_by(Id) %>%
mutate(Usage_Total = SedentaryMinutes + LightlyActiveMinutes +
FairlyActiveMinutes + VeryActiveMinutes) %>%
summarize(Usage_Total = mean(Usage_Total)) %>%
mutate(Usage_Total_Percentage = (Usage_Total/1440) * 100) %>% #There are 1440 minutes in 24 hours
mutate(Usage_Type = case_when(
Usage_Total_Percentage > 90 ~ 'More than 90% of the day',
Usage_Total_Percentage <= 90 & Usage_Total_Percentage >= 60 ~ 'Between 60-90% of the day'))
glimpse(daily_usage)## Rows: 33
## Columns: 4
## $ Id <dbl> 1503960366, 1624580081, 1644430081, 1844505072,…
## $ Usage_Total <dbl> 1125.9677, 1425.7097, 1371.2667, 1323.4839, 135…
## $ Usage_Total_Percentage <dbl> 78.19220, 99.00762, 95.22685, 91.90860, 94.3122…
## $ Usage_Type <chr> "Between 60-90% of the day", "More than 90% of …# Create table on percentage of user type grouped by 'Usage_Type'
daily_usage_user_type <- daily_usage %>%
group_by(Usage_Type) %>%
summarize(Usage_Type_Count = n()) %>%
mutate(Usage_Type_Percentage = paste0(round(Usage_Type_Count/nrow(daily_usage) * 100), '%'))
glimpse(daily_usage_user_type)## Rows: 2
## Columns: 3
## $ Usage_Type <chr> "Between 60-90% of the day", "More than 90% of t…
## $ Usage_Type_Count <int> 15, 18
## $ Usage_Type_Percentage <chr> "45%", "55%"# Create pie chart
ggplot(daily_usage_user_type, aes(x = '', y = Usage_Type_Count, fill = Usage_Type)) +
geom_bar(stat = 'identity', colour = 'black') +
coord_polar(theta = 'y') +
theme_void() +
geom_text(aes(label = Usage_Type_Percentage), position = position_stack(vjust = 0.5)) +
guides(fill = guide_legend(reverse = TRUE)) +
labs(title = 'Average Daily FitBit Usage', fill = 'FitBit Usage') +
theme(plot.title = element_text(size = 15, hjust = 0.5))The pie chart shows 55% of the FitBit users used their FitBit device more than 90% of the day, while the remaining users used the device between 60-90% of the day.
4.3.2 FitBit User Types on Monthly Usage
I created a table to determine the proportion of users that use the FitBit device Every day, more than half the month, and less than half the month.
# Count number of unique days in 'daily_activity'
Total_Days = daily_activity %>%
summarize(Total_Days = n_distinct(Date)) %>%
pull(Total_Days)
# Check data
Total_Days## [1] 31# Create table on average monthly usage of FitBit grouped by 'Id'
monthly_usage <- daily_activity %>%
group_by(Id) %>%
summarize(Usage_Days = n()) %>%
mutate(Usage_Days_percentage = Usage_Days/Total_Days*100) %>%
mutate(User_Type = case_when(
Usage_Days_percentage == 100 ~ 'Every day',
Usage_Days_percentage < 100 & Usage_Days_percentage >= 50 ~ 'More than half the month',
Usage_Days_percentage < 50 ~ 'Less than half the month'))
# Categorize the 'User_Type' in order
User_Type_Order <- c('Every day', 'More than half the month', 'Less than half the month')
monthly_usage$User_Type <- factor(monthly_usage$User_Type, levels = User_Type_Order)
# Check data
glimpse(monthly_usage)## Rows: 33
## Columns: 4
## $ Id <dbl> 1503960366, 1624580081, 1644430081, 1844505072, …
## $ Usage_Days <int> 31, 31, 30, 31, 31, 31, 31, 31, 18, 31, 20, 30, …
## $ Usage_Days_percentage <dbl> 100.00000, 100.00000, 96.77419, 100.00000, 100.0…
## $ User_Type <fct> Every day, Every day, More than half the month, …# Create table on of percentage of user type grouped by 'usage_type'
monthly_usage_percentage <- monthly_usage %>%
group_by(User_Type) %>%
summarize(User_Type_Count = n()) %>%
mutate(Percentage = paste0(round(User_Type_Count/nrow(monthly_usage)*100), '%'))
# Check data
glimpse(monthly_usage_percentage)## Rows: 3
## Columns: 3
## $ User_Type <fct> Every day, More than half the month, Less than half th…
## $ User_Type_Count <int> 21, 11, 1
## $ Percentage <chr> "64%", "33%", "3%"# Create pie chart
ggplot(monthly_usage_percentage, aes(x = '', y = User_Type_Count, fill = User_Type)) +
geom_bar(stat = 'identity', colour = 'black') +
coord_polar(theta = 'y', start = 1.273*pi) +
theme_void() +
geom_text(aes(label = Percentage), position = position_stack(vjust = 0.5)) +
guides(fill = guide_legend(reverse = FALSE)) +
labs(title = 'Average Monthly FitBit Usage', fill = 'FitBit Usage') +
theme(plot.title = element_text(size = 15, hjust = 0.5)) +
scale_fill_manual(values = c("Every day" = "turquoise1",
"More than half the month" = "springgreen4",
"Less than half the month" = "indianred2"))Nearly two-thirds of the users used their FitBit device every day while only 3% of the users used their device less than half of the month.
4.3.3 Daily Steps
I created a table on daily steps that filtered out the last date due to the unusually low records on that specific day, as mentioned in section 2.4.4 and shown on the chart.
# Grouped then filter out last date due to unusually low records that can skew data
daily_steps <- daily_activity %>%
group_by(Date) %>%
summarise(Avg_Total_Steps = mean(TotalSteps)) %>%
filter(Date != '2016-05-12')
# Check data
glimpse(daily_steps)## Rows: 30
## Columns: 2
## $ Date <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-15, 2016-…
## $ Avg_Total_Steps <dbl> 8236.848, 7198.727, 7743.576, 7533.848, 8679.156, 6409…# Store average steps in variable
mean_steps <- mean(daily_steps$Avg_Total_Steps)
mean_steps## [1] 7728.675# Create graph
ggplot(daily_steps, aes(x = Date, y = Avg_Total_Steps)) +
geom_bar(stat = 'identity', fill = 'chocolate', color = 'black', width = 1) +
labs(x = 'Date', y = 'Steps', title = 'Average Daily Steps') +
theme(axis.text.x = element_text(angle=15, vjust=1, hjust=1),
plot.title = element_text(hjust = 0.5)) +
geom_smooth(method= 'loess', se=FALSE, colour = 'chocolate4', linetype = 'dashed') +
geom_smooth(method= lm, se=FALSE, colour = 'darkblue') +
coord_cartesian(ylim = c(6000, 9000)) +
annotate('text', x = mean(daily_steps$Date), y = 8600,
label = paste('Average =', round(mean_steps), 'steps'), size = 4, colour = 'red')The graph shows a slight downtrend in average steps for the overall FitBit user population from this data alone. However, interestingly enough, the results will be the opposite when we filter the data to include only those who have also recorded their sleep duration. I discuss this in detail in section 4.8.2.
4.3.4 Daily Steps and Calories Burned From Users Without Sleep Data
I want to know how the data looks when filtering only data from users that did not record their sleep.
# Check for unique combinations of 'Time' and 'AM_or_PM' values
daily_sleep_Time_check <- unique(paste(daily_sleep$Time, daily_sleep$AM_or_PM))
daily_sleep_Time_check## [1] "12:00:00 AM"# Merged 'daily_activity' and 'daily_sleep' to filter data with activity and sleep records
daily_combined <- merge(daily_activity, daily_sleep, by = c('Date', 'Id'))
# Get the list of Ids that had records for both daily activities and sleep
daily_combined_unique_id <- daily_combined %>%
distinct(Id)
# Check data
glimpse(daily_combined_unique_id)## Rows: 24
## Columns: 1
## $ Id <dbl> 1503960366, 1927972279, 2026352035, 3977333714, 4020332650, 4445114…# Create table that filters out the Ids from 'daily_combined_unique_id'
daily_activity_ids_without_sleep_record <- daily_activity %>%
filter(!Id %in% daily_combined_unique_id) %>%
filter(Date != '2016-05-12') # Filter out last day '2016-05-12' due to outlier results
# Check data
glimpse(daily_activity_ids_without_sleep_record)## Rows: 919
## Columns: 15
## $ Id <dbl> 1503960366, 1503960366, 1503960366, 150396036…
## $ Date <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-…
## $ TotalSteps <dbl> 13162, 10735, 10460, 9762, 12669, 9705, 13019…
## $ TotalDistance <dbl> 8.50, 6.97, 6.74, 6.28, 8.16, 6.48, 8.59, 9.8…
## $ TrackerDistance <dbl> 8.50, 6.97, 6.74, 6.28, 8.16, 6.48, 8.59, 9.8…
## $ LoggedActivitiesDistance <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ VeryActiveDistance <dbl> 1.88, 1.57, 2.44, 2.14, 2.71, 3.19, 3.25, 3.5…
## $ ModeratelyActiveDistance <dbl> 0.55, 0.69, 0.40, 1.26, 0.41, 0.78, 0.64, 1.3…
## $ LightActiveDistance <dbl> 6.06, 4.71, 3.91, 2.83, 5.04, 2.51, 4.71, 5.0…
## $ SedentaryActiveDistance <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ VeryActiveMinutes <dbl> 25, 21, 30, 29, 36, 38, 42, 50, 28, 19, 66, 4…
## $ FairlyActiveMinutes <dbl> 13, 19, 11, 34, 10, 20, 16, 31, 12, 8, 27, 21…
## $ LightlyActiveMinutes <dbl> 328, 217, 181, 209, 221, 164, 233, 264, 205, …
## $ SedentaryMinutes <dbl> 728, 776, 1218, 726, 773, 539, 1149, 775, 818…
## $ Calories <dbl> 1985, 1797, 1776, 1745, 1863, 1728, 1921, 203…# Create new table that tracks average Steps, Calories, and Activity grouped by 'Date'
daily_activity_ids_without_sleep_record_avg <- daily_activity_ids_without_sleep_record %>%
group_by(Date) %>%
summarize(Steps = mean(TotalSteps), Calories = mean(Calories),
LightlyActiveMinutes = mean(LightlyActiveMinutes),
FairlyActiveMinutes = mean(FairlyActiveMinutes),
VeryActiveMinutes = mean(VeryActiveMinutes))
# Calculate the averages for Steps and Calories
Steps_Avg_ids_without_sleep_record <- mean(daily_activity_ids_without_sleep_record_avg$Steps)
Calories_Avg_ids_without_sleep_record <- mean(daily_activity_ids_without_sleep_record_avg$Calories)# Plot Steps graph
ggplot(daily_activity_ids_without_sleep_record_avg, aes(x = Date, y = Steps)) +
geom_bar(stat = 'identity', colour = 'black', fill = 'skyblue') +
geom_hline(yintercept = Steps_Avg_ids_without_sleep_record, linetype = 'dashed',
colour = 'red', linewidth = 0.75) +
geom_smooth(method = lm, se = FALSE, linewidth = 1.5, colour = 'darkblue') +
labs(title = 'Daily Steps', subtitle ='From FitBit Users Without Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5),
plot.subtitle = element_text(size = 8)) +
annotate('text', x = as.Date('2016-05-04'), y = 11500,
label = paste0('Average Steps: ', round(Steps_Avg_ids_without_sleep_record), '/day'),
colour = 'darkred') +
coord_cartesian(ylim = c(0, 11500)) +
scale_y_continuous(breaks = c(3000, 6000, 9000, 12000))There is a slight downtrend in steps taken throughout the month for FitBit users who did not record their sleep patterns at all.
4.3.5 Daily Activity Level Based on Activity Type
I created a table on daily activity levels that filtered out the last date since it has an unusually low amount of records compared to the average day. I also create a variable to store the average values of various activity types based on the date.
# Filter out last date due to unusually low records that can skew data
daily_active_minutes <- daily_activity %>%
group_by(Date) %>%
summarise(Lightly_Active_Minutes = mean(LightlyActiveMinutes),
Fairly_Active_Minutes = mean(FairlyActiveMinutes),
Very_Active_Minutes = mean(VeryActiveMinutes)) %>%
filter(Date != '2016-05-12')
# Check data
glimpse(daily_active_minutes)## Rows: 30
## Columns: 4
## $ Date <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-15…
## $ Lightly_Active_Minutes <dbl> 199.0000, 181.7576, 201.0000, 213.8485, 193.812…
## $ Fairly_Active_Minutes <dbl> 7.848485, 10.575758, 12.393939, 9.878788, 15.12…
## $ Very_Active_Minutes <dbl> 22.30303, 20.33333, 20.93939, 19.18182, 27.8437…# Store average values for various activitly levels in separate variables
mean_lightly_active <- mean(daily_active_minutes$Lightly_Active_Minutes)
mean_fairly_active <- mean(daily_active_minutes$Fairly_Active_Minutes)
mean_very_active <- mean(daily_active_minutes$Very_Active_Minutes)
# Check average values
mean_lightly_active## [1] 194.6625mean_fairly_active## [1] 13.90128mean_very_active## [1] 21.51954# Create graph
ggplot(daily_active_minutes, aes(x = Date)) +
geom_line(aes(y = Lightly_Active_Minutes, color = "Lightly Active")) +
geom_line(aes(y = Fairly_Active_Minutes, color = "Fairly Active")) +
geom_line(aes(y = Very_Active_Minutes, color = "Very Active")) +
labs(x = "Date", y = "Minutes", title = "Daily Activity", colour = "Activity") +
scale_color_manual(values = c("Lightly Active" = "dodgerblue", "Fairly Active" = "olivedrab", "Very Active" = "tomato1"),
breaks = c("Lightly Active", "Fairly Active", "Very Active")) +
geom_smooth(aes(y = Lightly_Active_Minutes), method = "loess", se = FALSE, color = "blue", linewidth = 1) +
geom_smooth(aes(y = Fairly_Active_Minutes), method = "loess", se = FALSE, color = "darkgreen", linewidth = 1) +
geom_smooth(aes(y = Very_Active_Minutes), method = "loess", se = FALSE, color = "red", linewidth = 1) +
theme(axis.text.x = element_text(angle = 15, vjust = 1, hjust = 1),
axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5),
legend.title = element_text(size = 15)) +
scale_y_continuous(breaks = seq(0, 250, by = 25)) +
annotate("text", x = as.Date("2016-05-07"), y = 225, label = 'Average = 194', colour = "blue") +
annotate("text", x = as.Date("2016-05-07"), y = 65, label = 'Average = 21', colour = "red") +
annotate("text", x = as.Date("2016-05-07"), y = 50, label = 'Average = 13', colour = "darkgreen")While there is no clear trend in the fairly active and very active levels, there is a slight downtrend, in the light activity level.
4.3.6 Daily Calories
I created a table on daily calories grouped by date. Again, the last date was filtered out.
# Create table on daily calories
daily_calories<- daily_activity %>%
group_by(Date) %>%
summarise(Calories_Burned = mean(Calories)) %>%
filter(Date != '2016-05-12')
# Check data
glimpse(daily_calories)## Rows: 30
## Columns: 2
## $ Date <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-15, 2016-…
## $ Calories_Burned <dbl> 2390.697, 2286.636, 2356.394, 2355.182, 2392.938, 2230…# Create variable to store average calories
mean_calories <- mean(daily_calories$Calories_Burned)
mean_calories## [1] 2328.284ggplot(daily_calories, aes(x = Date, y = Calories_Burned)) +
geom_bar(stat = "identity", fill = "skyblue", color = "black", width = 1) +
labs(x = "Date", y = "Calories", title = "Average Daily Calories Burned") +
theme(axis.text.x = element_text(angle = 15, vjust = 1, hjust = 1),
axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5),
legend.title = element_text(size = 15)) +
geom_smooth(method = lm, se = FALSE, colour = 'darkblue') +
annotate("text", x = mean(daily_calories$Date), y = 2550, label = "Average = 2328 calories", size = 4, colour = "red")There also seems to be a slight decrease in calories burned over time.
Key Findings
Nearly two-thirds of the FitBit users consistently used their FitBit device every day and more than half of all users use their device throughout 90% of the day
There is a slight downtrend in steps taken, light activity, and calories burned throughout the one-month record
4.4 Analyzing ‘hourlyCalories_merged_cleaned’
I created a table on hourly calories then grouped the data by time. I also created variables to store the highlighted areas on the next graph.
# Create a new column in datetime 24h format
hourly_calories$Time_24h <- as.POSIXct(paste(hourly_calories$Time, hourly_calories$AM_or_PM), format = '%I:%M:%S %p')
glimpse(hourly_calories)## Rows: 22,099
## Columns: 6
## $ Id <dbl> 1503960366, 1503960366, 1503960366, 1503960366, 1503960366, 1…
## $ Date <date> 2016-04-12, 2016-04-12, 2016-04-12, 2016-04-12, 2016-04-12, …
## $ Time <time> 12:00:00, 01:00:00, 02:00:00, 03:00:00, 04:00:00, 05:00:00, …
## $ AM_or_PM <chr> "AM", "AM", "AM", "AM", "AM", "AM", "AM", "AM", "AM", "AM", "…
## $ Calories <dbl> 81, 61, 59, 47, 48, 48, 48, 47, 68, 141, 99, 76, 73, 66, 110,…
## $ Time_24h <dttm> 2024-06-05 00:00:00, 2024-06-05 01:00:00, 2024-06-05 02:00:0…# Get current date
current_date <- Sys.Date()
# Highlighted areas on Calories 24h graph
highlight3 <- data.frame(xmin = as.POSIXct(paste(current_date, '04:30:00')),
xmax = as.POSIXct(paste(current_date, '06:30:00')),
ymin = 63, ymax = Inf)
highlight4 <- data.frame(xmin = as.POSIXct(paste(current_date, '11:30:00')),
xmax = as.POSIXct(paste(current_date, '19:30:00')),
ymin = 63, ymax = Inf)# Plot Graph
ggplot(hourly_calories, aes(x = Time_24h, y = Calories)) +
geom_rect(data = highlight3, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'lightcoral', alpha = 0.5, inherit.aes = FALSE) +
geom_rect(data = highlight4, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'lightcoral', alpha = 0.5, inherit.aes = FALSE) +
geom_point(colour = 'darkblue', alpha = 0.6) +
geom_jitter(colour = 'darkblue', alpha = 0.1) +
geom_smooth(method = 'loess', span = 0.1, se = FALSE) +
theme(axis.text.x = element_blank(), axis.title.x = element_text(margin = margin(t = 30))) +
annotate('text', x = as.POSIXct(paste(current_date, '00:00:00')), y = -100, label = '00:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '06:00:00')), y = -100, label = '06:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:00:00')), y = -100, label = '12:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '18:00:00')), y = -100, label = '18:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '23:59:00')), y = -100, label = '23:59h') +
annotate('text', x = as.POSIXct(paste(current_date, '05:30:00')), y = -10, label = '(05:00-07:00h)',
colour = 'darkred') +
annotate('text', x = as.POSIXct(paste(current_date, '15:30:00')), y = -10, label = '(12:00-20:00h)',
colour = 'darkred') +
coord_cartesian(ylim = c(0, 1000), clip = 'off') +
labs(x = 'Time of Day', y = 'Calories Burned', title = 'Instances of Calories Burned Hourly') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5)) The highlighted regions on the graph indicate the typical times of day when the highest calorie burn occurs. Because the data on time is measured strictly by the hour on the dot (Ex: 12:00am, 1:00am, 2:00am, etc.), I added jitter to the plots which results in slight variations of positions to the plots that overlap. This way, it is easier to visualize the areas with the greatest intensity (recorded points).
The blue trend line show a slight increase in average calories burned starting from 5am and lasting until 8pm.
## Warning in simpleLoess(y, x, w, span, degree = degree, parametric = parametric,
## : pseudoinverse used at 1.7176e+09## Warning in simpleLoess(y, x, w, span, degree = degree, parametric = parametric,
## : neighborhood radius 7614## Warning in simpleLoess(y, x, w, span, degree = degree, parametric = parametric,
## : reciprocal condition number 3.7961e-15## Warning in simpleLoess(y, x, w, span, degree = degree, parametric = parametric,
## : There are other near singularities as well. 5.7973e+07
Key Findings
- Calorie burns higher throughout the day between 5am and 8pm, with instances of the highest calories burned occurring between 12pm-8pm
4.5 Analyzing ‘hourlyIntensities_merged_cleaned’
I created a table on the hourly intensities grouped by time.
# Create a new column in datetime 24h format
hourly_intensities$Time_24h <- as.POSIXct(paste(hourly_intensities$Time, hourly_intensities$AM_or_PM), format = "%I:%M:%S %p")
# Get current date
current_date <- Sys.Date()
# Summarize data
hourly_intensities_avg <- hourly_intensities %>%
group_by(Time_24h) %>%
summarize(Intensity_Average = mean(TotalIntensity))
# Check data
glimpse(hourly_intensities_avg)## Rows: 24
## Columns: 2
## $ Time_24h <dttm> 2024-06-05 00:00:00, 2024-06-05 01:00:00, 2024-06-0…
## $ Intensity_Average <dbl> 2.1295503, 1.4190782, 1.0439443, 0.4437299, 0.633047…# Calculate the value at the peaks
hourly_intensities_at_12h_18h <- hourly_intensities_avg %>%
filter(Time_24h %in% as.POSIXct(paste(current_date, c('12:00:00', '18:00:00'))))
# Check data
glimpse(hourly_intensities_at_12h_18h)## Rows: 2
## Columns: 2
## $ Time_24h <dttm> 2024-06-05 12:00:00, 2024-06-05 18:00:00
## $ Intensity_Average <dbl> 19.84707, 21.92163# Rounding the values at the peaks
round(hourly_intensities_at_12h_18h$Intensity_Average[1], 1)## [1] 19.8round(hourly_intensities_at_12h_18h$Intensity_Average[2], 1)## [1] 21.9# Plotting the graph
ggplot(hourly_intensities_avg, aes(x = Time_24h, y = Intensity_Average)) +
geom_area(colour = 'black', fill = 'skyblue', alpha = 0.5) +
theme(axis.text.x = element_blank(), axis.title.x = element_text(margin = margin(t = 30))) +
annotate('text', x = as.POSIXct(paste(current_date, '00:00:00')), y = -2.5, label = '00:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '06:00:00')), y = -2.5, label = '06:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:00:00')), y = -2.5, label = '12:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '18:00:00')), y = -2.5, label = '18:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '23:59:00')), y = -2.5, label = '23:59h') +
coord_cartesian(ylim = c(0, 25), clip = 'off') +
labs(x = 'Time of Day', y = 'Intensity', title = 'Hourly Average Intensity') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5)) +
geom_point(data = hourly_intensities_at_12h_18h, aes(x = Time_24h, y = Intensity_Average),
colour = 'darkred') +
annotate('text', x = as.POSIXct(hourly_intensities_at_12h_18h$Time_24h[1]),
y = hourly_intensities_at_12h_18h$Intensity_Average[1] + 1,
label = paste(round(hourly_intensities_at_12h_18h$Intensity_Average[1], 1))) +
annotate('text', x = as.POSIXct(hourly_intensities_at_12h_18h$Time_24h[2]),
y = hourly_intensities_at_12h_18h$Intensity_Average[2] + 1,
label = paste(round(hourly_intensities_at_12h_18h$Intensity_Average[2], 1)))While it is ambiguous what the value of intensity represents (see section 2.3), it is clear that the level of intensity peaks at 12:00h and 18:00h. Perhaps the increased intensity level at noon is due to people going for a walk on their way to grab lunch during working days. The peak intensity level in the evening may be due to activities after work such as running errands, or perhaps this is the peak time of exercise for most people.
Key Findings
- Intensities are the highest at 12pm and 6pm
4.6 Analyzing ‘’hourlySteps_merged_cleaned’
I created a table on hourly steps grouped according to time. I also created two variables to store the data for the highlighted areas on the next graph.
# Create a new column in datetime 24h format
hourly_steps$Time_24h <- as.POSIXct(paste(hourly_steps$Time, hourly_steps$AM_or_PM), format = '%I:%M:%S %p')
# Get current date
current_date <- Sys.Date()
# Summarize data
hourly_steps_avg <- hourly_steps %>%
group_by(Time_24h) %>%
summarize(Steps_Average = mean(StepTotal))
glimpse(hourly_steps_avg)## Rows: 24
## Columns: 2
## $ Time_24h <dttm> 2024-06-05 00:00:00, 2024-06-05 01:00:00, 2024-06-05 02…
## $ Steps_Average <dbl> 42.188437, 23.102894, 17.110397, 6.426581, 12.699571, 43…#Highlighted areas on Calories 24h graph
highlight5 <- data.frame(xmin = as.POSIXct(paste(current_date, '11:45:00')),
xmax = as.POSIXct(paste(current_date, '14:30:00')),
ymin = 0, ymax = Inf)
highlight6 <- data.frame(xmin = as.POSIXct(paste(current_date, '17:30:00')),
xmax = as.POSIXct(paste(current_date, '19:30:00')),
ymin = 0, ymax = Inf)# Plotting the graph
ggplot(hourly_steps_avg, aes(x = Time_24h, y = Steps_Average)) +
geom_rect(data = highlight5, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = "lightcoral", alpha = 0.5, inherit.aes = FALSE) +
geom_rect(data = highlight6, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = "lightcoral", alpha = 0.5, inherit.aes = FALSE) +
geom_area(colour = 'black', fill = 'skyblue', alpha = 0.5) +
theme(axis.text.x = element_blank(), axis.title.x = element_text(margin = margin(t = 30))) +
annotate('text', x = as.POSIXct(paste(current_date, '00:00:00')), y = -75, label = '00:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '06:00:00')), y = -75, label = '06:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:00:00')), y = -75, label = '12:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '18:00:00')), y = -75, label = '18:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '23:59:00')), y = -75, label = '23:59h') +
annotate("text", x = as.POSIXct(paste(current_date, '13:00:00')), y = -13, label = "(12:00-14:00h)",
colour = 'darkred') +
annotate('text', x = as.POSIXct(paste(current_date, '18:30:00')), y = -12, label = '(18:00-19:00h)',
colour = 'darkred') +
coord_cartesian(ylim = c(0, 650), clip = 'off') +
labs(x = 'Time of Day', y = 'Steps', title = 'Hourly Average Steps') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5))The amount of steps taken by the hour peaks at similar time frames as the peak calories burned and intensity discussed in section 4.4 and section 4.5, respectively. Based on the data, it is probable that steps influence the intensity levels of the day and the calories burned. At the very least, there is a correlation between these data.
Key Findings
- FitBit Users walk the most around lunch (12-2pm) and dinner (6-7pm) time
4.7 Analyzing ‘minuteSleep_merged_cleaned’
I created a table on sleep time grouped by the minute. I also create two variables that store the information for the highlighted areas on the next graph.
# Create a new column in datetime 24h format
minutes_sleep$Time_24h <- as.POSIXct(paste(minutes_sleep$Time, minutes_sleep$AM_or_PM), format = '%I:%M:%S %p')
# Get current date
current_date <- Sys.Date()
# Highlighted areas on Calories 24h graph
highlight5 <- data.frame(xmin = as.POSIXct(paste(current_date, '00:00:00')),
xmax = as.POSIXct(paste(current_date, '11:30:00')),
ymin = 0, ymax = Inf)
highlight6 <- data.frame(xmin = as.POSIXct(paste(current_date, '20:00:00')),
xmax = as.POSIXct(paste(current_date, '23:59:00')),
ymin = 0, ymax = Inf)# Plotting the graph
ggplot(minutes_sleep, aes(x = Time_24h)) +
geom_rect(data = highlight5, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'royalblue4', alpha = 0.3, inherit.aes = FALSE) +
geom_rect(data = highlight6, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'royalblue4', alpha = 0.3, inherit.aes = FALSE) +
geom_area(colour = 'slateblue4', fill = 'slateblue4', stat = 'bin', binwidth = 360, alpha = 0.5) +
labs(x = 'Time of Day', y = 'Relative Sleep Sessions', title = 'Sleep Sessions Distribution') +
theme(axis.text.x = element_blank(), axis.text.y = element_blank(), axis.ticks.y = element_blank(),
axis.title.x = element_text(margin = margin(t = 30))) +
annotate('text', x = as.POSIXct(paste(current_date, '00:00:00')), y = -225, label = '00:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '06:00:00')), y = -225, label = '06:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:00:00')), y = -225, label = '12:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '18:00:00')), y = -225, label = '18:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '23:59:00')), y = -225, label = '23:59h') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5)) +
coord_cartesian(ylim = c(0, 2300), clip = "off") +
annotate('text', x = as.POSIXct(paste(current_date, '10:15:00')), y = 1750, label = '11:30h') +
annotate('text', x = as.POSIXct(paste(current_date, '21:15:00')), y = 1750, label = '20:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '07:00:00')), y = 2125, label = '05:30h',
colour = 'darkred') +
geom_vline(xintercept = as.POSIXct(paste(current_date, '11:30:00')),
linetype = 'dashed') +
geom_vline(xintercept = as.POSIXct(paste(current_date, '20:00:00')),
linetype = 'dashed') +
geom_vline(xintercept = as.POSIXct(paste(current_date, '05:30:00')),
linetype = 'dashed', colour = 'red') +
annotate('text', x = as.POSIXct(paste(current_date, '22:30:00')), y = 2125,
label = '\U0001F6CF', size = 8)While there are records of sleep data throughout the whole day, it is clear that the there is a sharp increase in sleep sessions starting 8pm and peaks around 4am. At 5:30am, there is a sharp decrease in sleep session until 11:30am. This means that many people start going to bed by 8pm and the majority of people would have already been awake by 11:30am.
Key Findings
- Most FitBit users start going to bed by 8pm and the majority are already awake by 11:30am
4.8 Analyzing ‘sleepDay_merged_cleaned’
4.8.1 Merging tables ‘daily_activity’ with ‘daily_sleep’
I first created a table on sleep time then created another table that is a merged version of the sleep table and the daily activity table from section 4.3. This means that only data by users that includes records from both sleep time and daily activity will be selected. If a user records data on sleep but not daily activity on a particular day (or vice versa), the data will be not be included. This merged table provides a deeper analysis on the correlation between daily activity and sleep time on the same users.
# Check for unique combinations of 'Time' and 'AM_or_PM' values (from 4.3.4)
daily_sleep_Time_check <- unique(paste(daily_sleep$Time, daily_sleep$AM_or_PM))
daily_sleep_Time_check## [1] "12:00:00 AM"# Merged 'daily_activity' and 'daily_sleep' to check for correlation (from 4.3.4)
daily_combined <- merge(daily_activity, daily_sleep, by = c('Date', 'Id'))
# Filter out last day '2016-05-12' due to outlier results (low activity due to last day?)
daily_combined <- daily_combined %>%
filter(Date != '2016-05-12')
# Add a new column for the time difference between 'Time_In_Bed' and 'Minutes_Asleep'
daily_combined$Time_In_Bed_Awake <- daily_combined$TotalTimeInBed - daily_combined$TotalMinutesAsleep
# Convert to Day_Of_Week with 'ordered' data type
weekday_order <- c('Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun')
daily_combined$Day_Of_Week <- substr(format(daily_combined$Date, '%a'), 1, 3)
daily_combined$Day_Of_Week <- ordered(daily_combined$Day_Of_Week, levels = weekday_order)
# Check that the data makes sense
glimpse(daily_combined)## Rows: 402
## Columns: 22
## $ Date <date> 2016-04-12, 2016-04-12, 2016-04-12, 2016-04-…
## $ Id <dbl> 1503960366, 1927972279, 2026352035, 397733371…
## $ TotalSteps <dbl> 13162, 678, 4414, 8856, 8539, 3276, 7213, 115…
## $ TotalDistance <dbl> 8.50, 0.47, 2.74, 5.98, 6.12, 2.20, 5.88, 7.5…
## $ TrackerDistance <dbl> 8.50, 0.47, 2.74, 5.98, 6.12, 2.20, 5.88, 7.5…
## $ LoggedActivitiesDistance <dbl> 0.000000, 0.000000, 0.000000, 0.000000, 0.000…
## $ VeryActiveDistance <dbl> 1.88, 0.00, 0.19, 3.06, 0.15, 0.00, 0.00, 1.3…
## $ ModeratelyActiveDistance <dbl> 0.55, 0.00, 0.35, 0.91, 0.24, 0.00, 0.00, 0.7…
## $ LightActiveDistance <dbl> 6.06, 0.47, 2.20, 2.01, 5.68, 2.20, 5.85, 5.4…
## $ SedentaryActiveDistance <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ VeryActiveMinutes <dbl> 25, 0, 3, 44, 4, 0, 0, 19, 86, 50, 59, 65, 0,…
## $ FairlyActiveMinutes <dbl> 13, 0, 8, 19, 15, 0, 0, 13, 16, 14, 6, 15, 0,…
## $ LightlyActiveMinutes <dbl> 328, 55, 181, 131, 331, 196, 263, 277, 140, 1…
## $ SedentaryMinutes <dbl> 728, 734, 706, 777, 712, 787, 718, 767, 728, …
## $ Calories <dbl> 1985, 2220, 1459, 1450, 3654, 2113, 2947, 202…
## $ Time <time> 12:00:00, 12:00:00, 12:00:00, 12:00:00, 12:0…
## $ AM_or_PM <chr> "AM", "AM", "AM", "AM", "AM", "AM", "AM", "AM…
## $ TotalSleepRecords <dbl> 1, 3, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 2, 1, …
## $ TotalMinutesAsleep <dbl> 327, 750, 503, 274, 501, 429, 425, 441, 419, …
## $ TotalTimeInBed <dbl> 346, 775, 546, 469, 541, 457, 439, 464, 438, …
## $ Time_In_Bed_Awake <dbl> 19, 25, 43, 195, 40, 28, 14, 23, 19, 21, 11, …
## $ Day_Of_Week <ord> Tue, Tue, Tue, Tue, Tue, Tue, Tue, Tue, Tue, …# Summarize the data by date
daily_combined_avg_by_date <- daily_combined %>%
group_by(Date) %>%
summarize(Time_In_bed = mean(TotalTimeInBed), Minutes_Asleep = mean(TotalMinutesAsleep),
Calories = mean(Calories), Sedentary_Minutes = mean(SedentaryMinutes),
Lightly_Active_Minutes = mean(LightlyActiveMinutes),
Fairly_Active_Minutes = mean(FairlyActiveMinutes),
Very_Active_Minutes = mean(VeryActiveMinutes),
Steps = mean(TotalSteps), Distance = mean(TotalDistance),
Time_In_Bed_Awake = mean(Time_In_Bed_Awake))
# Check the new table to see if it makes sense
glimpse(daily_combined_avg_by_date)## Rows: 30
## Columns: 11
## $ Date <date> 2016-04-12, 2016-04-13, 2016-04-14, 2016-04-15…
## $ Time_In_bed <dbl> 479.6923, 471.8571, 480.2308, 476.3529, 433.000…
## $ Minutes_Asleep <dbl> 441.9231, 430.4286, 445.2308, 427.4706, 391.714…
## $ Calories <dbl> 2438.769, 2207.929, 2425.615, 2317.118, 2494.42…
## $ Sedentary_Minutes <dbl> 750.1538, 766.1429, 742.7692, 777.4118, 710.357…
## $ Lightly_Active_Minutes <dbl> 193.5385, 163.5714, 194.4615, 199.5294, 226.785…
## $ Fairly_Active_Minutes <dbl> 9.153846, 13.571429, 17.538462, 14.352941, 17.2…
## $ Very_Active_Minutes <dbl> 27.30769, 18.42857, 27.53846, 18.17647, 31.8571…
## $ Steps <dbl> 7505.769, 6103.000, 7626.154, 7472.059, 8615.07…
## $ Distance <dbl> 5.260000, 4.249286, 5.348462, 5.221176, 6.15357…
## $ Time_In_Bed_Awake <dbl> 37.76923, 41.42857, 35.00000, 48.88235, 41.2857…# Summarize the data by day of week
daily_combined_avg_by_day_of_week <- daily_combined %>%
group_by(Day_Of_Week) %>%
summarize(Time_In_bed = mean(TotalTimeInBed), Minutes_Asleep = mean(TotalMinutesAsleep),
Calories = mean(Calories), Sedentary_Minutes = mean(SedentaryMinutes),
Lightly_Active_Minutes = mean(LightlyActiveMinutes),
Fairly_Active_Minutes = mean(FairlyActiveMinutes),
Very_Active_Minutes = mean(VeryActiveMinutes),
Steps = mean(TotalSteps), Distance = mean(TotalDistance),
Time_In_Bed_Awake = mean(Time_In_Bed_Awake))
# Check the new table to see if it makes sense
glimpse(daily_combined_avg_by_day_of_week)## Rows: 7
## Columns: 11
## $ Day_Of_Week <ord> Mon, Tue, Wed, Thu, Fri, Sat, Sun
## $ Time_In_bed <dbl> 457.3478, 443.2923, 470.0303, 429.8214, 445.052…
## $ Minutes_Asleep <dbl> 419.5000, 404.5385, 434.6818, 395.4286, 405.421…
## $ Calories <dbl> 2431.978, 2496.200, 2378.242, 2476.964, 2329.64…
## $ Sedentary_Minutes <dbl> 718.4130, 740.0462, 714.4545, 761.0714, 743.087…
## $ Lightly_Active_Minutes <dbl> 222.3696, 216.5846, 207.9697, 217.9286, 222.684…
## $ Fairly_Active_Minutes <dbl> 19.08696, 20.04615, 16.74242, 17.69643, 14.5789…
## $ Very_Active_Minutes <dbl> 30.71739, 30.61538, 21.33333, 25.75000, 21.1578…
## $ Steps <dbl> 9273.217, 9182.692, 8022.864, 8861.768, 7901.40…
## $ Distance <dbl> 6.540870, 6.430000, 5.719849, 6.261429, 5.51193…
## $ Time_In_Bed_Awake <dbl> 37.84783, 38.75385, 35.34848, 34.39286, 39.6315…# Calculate the mean for 'daily_combined_avg_by_day_of_week'
Sleep_Avg_Day_Of_Week <- mean(daily_combined_avg_by_day_of_week$Minutes_Asleep)
Steps_Avg_Day_Of_Week <- mean(daily_combined_avg_by_day_of_week$Steps)
Calories_Avg_Day_Of_Week <- mean(daily_combined_avg_by_day_of_week$Calories)
# Calculate the mean for 'daily_combined_avg_by_date'
Sleep_Avg_Date <- mean(daily_combined_avg_by_date$Minutes_Asleep)
Steps_Avg_Date <- mean(daily_combined_avg_by_date$Steps)
Calories_Avg_Date <- mean(daily_combined_avg_by_date$Calories)4.8.2 Plotting the steps graph from the merged table
# Plotting the steps graph by date
ggplot(daily_combined_avg_by_date, aes(x = Date, y = Steps)) +
geom_bar(stat = 'Identity', colour = 'black', fill = 'skyblue') +
geom_hline(yintercept = Steps_Avg_Date, linetype = 'dashed', colour = 'red',
linewidth = 0.75) +
geom_smooth(method = lm, se = FALSE, linewidth = 1.5, colour = 'darkblue') +
labs(title = 'Daily Steps', subtitle ='From FitBit Users With Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5),
plot.subtitle = element_text(size = 8)) +
annotate('text', x = as.Date('2016-05-04'), y = 11500,
label = paste0('Average Steps: ', round(Steps_Avg_Date), '/day'),
colour = 'darkred')In contrast to section 4.3.3, the daily steps of users who also recorded their sleep time show an uptrend in steps activity as the day passes. Does this suggest that users who are more dedicated in wearing their FitBit to record their activity make more effort in performing light activity? This could also just be a coincidence though since the merging of data omitted a large amount of records.
Key Findings
- FitBit Users who had records of both steps and sleep show an uptrend in steps taken
- These same users take about 8600 steps per day on average
4.8.3 Plotting the activity graph from the merged table
# Plotting the Activity graph
ggplot(daily_combined_avg_by_date, aes(x = Date)) +
geom_line(aes(y = Lightly_Active_Minutes), colour = 'skyblue') +
geom_line(aes(y = Fairly_Active_Minutes), colour = 'green') +
geom_line(aes(y = Very_Active_Minutes), colour = 'red') +
geom_line(aes(y = Minutes_Asleep), colour = 'violet') +
geom_smooth(aes(y = Lightly_Active_Minutes), method = lm, se = FALSE, colour = 'blue') +
geom_smooth(aes(y = Fairly_Active_Minutes), method = lm, se = FALSE, colour = 'darkgreen') +
geom_smooth(aes(y = Very_Active_Minutes), method = lm, se = FALSE, colour = 'darkred') +
geom_smooth(aes(y = Minutes_Asleep), method = lm, se = FALSE, colour = 'orchid4') +
labs(y = 'Minutes', title = 'Daily Activity',
subtitle ='From FitBit Users With Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5)) +
annotate('text', x = as.Date('2016-05-05'), y = 480, label = 'Sleep',
size = 5, colour = 'orchid4') +
annotate('text', x = as.Date('2016-05-05'), y = 300, label = 'Lightly Active',
size = 5, colour = 'blue') +
annotate('text', x = as.Date('2016-05-05'), y = 60, label = 'Fairly Active',
size = 5, colour = 'darkgreen') +
annotate('text', x = as.Date('2016-05-05'), y = 100, label = 'Very Active',
size = 5, colour = 'darkred') While the “Very Active” and “Fairly Active” levels remain constant, as light activity increases, sleep duration decreases. This could suggest that an increase in light activity decreases sleep quality. However, this seems unlikely. It is more probable that the increased light activity took up more of a user’s time which offered less time for sleep. If this is the case, it may be better to use time spent on light activity to perform higher activies for increased benefit.
Key Findings
- Sleep duration decreases as light activity increases
4.8.4 Plotting the steps graph by day of week from the merged table
# Plotting the steps graph
ggplot(daily_combined_avg_by_day_of_week, aes(x = Day_Of_Week, y = Steps)) +
geom_bar(stat = 'Identity', colour = 'black', fill = 'skyblue') +
geom_hline(yintercept = Steps_Avg_Day_Of_Week, linetype = 'dashed', colour = 'darkred',
linewidth = 0.75) +
annotate('text', x = 'Wed', y = 9800, colour = 'darkred',
label = paste0('Average Steps: ', round(Steps_Avg_Day_Of_Week), '/day')) +
annotate('text', x = 'Mon', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[1]))) +
annotate('text', x = 'Tue', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[2]))) +
annotate('text', x = 'Wed', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[3]))) +
annotate('text', x = 'Thu', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[4]))) +
annotate('text', x = 'Fri', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[5]))) +
annotate('text', x = 'Sat', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[6]))) +
annotate('text', x = 'Sun', y = -250,
label = paste(round(daily_combined_avg_by_day_of_week$Steps[7]))) +
labs(x = 'Day of Week', title = 'Steps by Day of the Week',
subtitle ='From FitBit Users With Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5))Users meet the recommended 7000-10000 steps per day, with the highest steps falling on a Saturday. This may be due to users running errands on this day, or doing other activities like hanging out with friends. In contrast, Sunday records the least amount of steps activity. Users can use this day to increase their physical activity for a more balanced lifestyle.
Key Findings
- FitBit Users take the most steps on Saturdays and the least steps on Sundays
- FitBit Users meet the recommended 7000-10000 steps per day
4.8.5 Plotting calories by day of week from merged table
# Plotting the calories graph by day of week
ggplot(daily_combined_avg_by_day_of_week, aes(x = Day_Of_Week, y = Calories)) +
geom_bar(stat = 'Identity', colour = 'black', fill = 'lightgoldenrod3') +
geom_hline(yintercept = Calories_Avg_Day_Of_Week, linetype = 'dashed', colour = 'darkred',
linewidth = 0.75) +
annotate('text', x = 'Wed', y = 2600, colour = 'darkred',
label = paste0('Average Calories: ', round(Calories_Avg_Day_Of_Week), '/day')) +
annotate('text', x = 'Mon', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[1]))) +
annotate('text', x = 'Tue', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[2]))) +
annotate('text', x = 'Wed', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[3]))) +
annotate('text', x = 'Thu', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[4]))) +
annotate('text', x = 'Fri', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[5]))) +
annotate('text', x = 'Sat', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[6]))) +
annotate('text', x = 'Sun', y = -90,
label = paste(round(daily_combined_avg_by_day_of_week$Calories[7]))) +
labs(x = 'Day of Week', title = 'Calories Burned by Day of the Week',
subtitle ='From FitBit Users With Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5))The average calories burned per day is 2414 with the least amount of calories burned on Sundays.
Key Findings
- FitBit Users burned an average of 2414 calories per day
4.8.6 Plotting the sleep time graph from the merged table
# Setting up for sleep time graph
Sleep_Avg_Check <- ifelse(daily_combined_avg_by_day_of_week$Minutes_Asleep > Sleep_Avg_Day_Of_Week,
"Over_Avg", "Below_Avg")# Plotting the sleep time graph by day of week
ggplot(daily_combined_avg_by_day_of_week, aes(x = Day_Of_Week, y = Minutes_Asleep)) +
geom_bar(stat = 'Identity', colour = 'black',
aes(fill = Sleep_Avg_Check)) +
scale_fill_manual(values = c("Over_Avg" = "green", "Below_Avg" = "orangered3")) +
geom_hline(yintercept = Sleep_Avg_Day_Of_Week, linetype = 'dashed', colour = 'darkred',
linewidth = 0.75) +
annotate('text', x = 'Wed', y = 500, colour = 'darkred',
label = paste0('Average Sleep: ', round(Sleep_Avg_Day_Of_Week), ' minutes/day')) +
annotate('text', x = 'Mon', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[1]))) +
annotate('text', x = 'Tue', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[2]))) +
annotate('text', x = 'Wed', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[3]))) +
annotate('text', x = 'Thu', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[4]))) +
annotate('text', x = 'Fri', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[5]))) +
annotate('text', x = 'Sat', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[6]))) +
annotate('text', x = 'Sun', y = -15,
label = paste(round(daily_combined_avg_by_day_of_week$Minutes_Asleep[7]))) +
labs(x = 'Day of Week', y = 'Minutes Asleep',
title = 'Daily Sleep Minutes by Day of the Week',
subtitle ='From FitBit Users With Sleep Records') +
theme(plot.title = element_text(size = 15, hjust = 0.5)) +
guides(fill = 'none')Users get 7 hours of sleep daily on average, with the most sleep falling on Sundays and the least sleep on Thursdays.
Key Findings
- FitBit Users sleep the most Sundays and the least on Thursdays
- The average sleep duration is about 7 hours
4.9 Analyzing ‘weightLogInfo_merged_cleaned.csv’
I created a table that filtered out IDs that only provided 2 or less records. 5 unique IDs and 63 records remain down from 8 unique IDs and 67 records
# Check data
glimpse(weight_log)## Rows: 67
## Columns: 10
## $ Id <dbl> 1503960366, 1503960366, 1927972279, 2873212765, 2873212…
## $ Date <date> 2016-05-02, 2016-05-03, 2016-04-13, 2016-04-21, 2016-0…
## $ Time <time> 11:59:59, 11:59:59, 01:08:52, 11:59:59, 11:59:59, 11:5…
## $ AM_or_PM <chr> "PM", "PM", "AM", "PM", "PM", "PM", "PM", "PM", "PM", "…
## $ WeightKg <dbl> 52.6, 52.6, 133.5, 56.7, 57.3, 72.4, 72.3, 69.7, 70.3, …
## $ WeightPounds <dbl> 115.9631, 115.9631, 294.3171, 125.0021, 126.3249, 159.6…
## $ Fat <dbl> 22, NA, NA, NA, NA, 25, NA, NA, NA, NA, NA, NA, NA, NA,…
## $ BMI <dbl> 22.65, 22.65, 47.54, 21.45, 21.69, 27.45, 27.38, 27.25,…
## $ IsManualReport <lgl> TRUE, TRUE, FALSE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, …
## $ LogId <dbl> 1.46e+12, 1.46e+12, 1.46e+12, 1.46e+12, 1.46e+12, 1.46e…# Filtered out Ids that provided 2 or less records
weight_log_filtered <- weight_log %>%
filter(!Id %in% c('1927972279', '5577150313', '1503960366'))
# Check filtered data
glimpse(weight_log_filtered)## Rows: 63
## Columns: 10
## $ Id <dbl> 2873212765, 2873212765, 4319703577, 4319703577, 4558609…
## $ Date <date> 2016-04-21, 2016-05-12, 2016-04-17, 2016-05-04, 2016-0…
## $ Time <time> 11:59:59, 11:59:59, 11:59:59, 11:59:59, 11:59:59, 11:5…
## $ AM_or_PM <chr> "PM", "PM", "PM", "PM", "PM", "PM", "PM", "PM", "PM", "…
## $ WeightKg <dbl> 56.7, 57.3, 72.4, 72.3, 69.7, 70.3, 69.9, 69.2, 69.1, 6…
## $ WeightPounds <dbl> 125.0021, 126.3249, 159.6147, 159.3942, 153.6622, 154.9…
## $ Fat <dbl> NA, NA, 25, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA,…
## $ BMI <dbl> 21.45, 21.69, 27.45, 27.38, 27.25, 27.46, 27.32, 27.04,…
## $ IsManualReport <lgl> TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, T…
## $ LogId <dbl> 1.46e+12, 1.46e+12, 1.46e+12, 1.46e+12, 1.46e+12, 1.46e…# Plotting the graph
ggplot(weight_log_filtered, aes(x = Date, y = WeightKg, colour = factor(Id))) +
geom_line() +
theme_minimal() +
labs(x = 'Date', y = 'Weight (Kg)', title = 'Daily Change in Weight (Kg)') +
theme(plot.title = element_text(size = 15, hjust = 0.5)) +
guides(colour = 'none')Each coloured line represents a user, with only 5 represented in this graph. There are no noticeable trends from the little data I analyzed.
Key Findings
- There is no clear trend in weight change for the 5 FitBit users that recorded their weight
4.10 Analyzing ‘heartrate_seconds_merged_cleaned’
4.10.1 Plotting the graph for the hourly average heart rate
I created a table to track heart rate according to the time of day. I also created variables to store the highlighted areas on the following graph.
# Create a new column in datetime 24h format
heart_rate$Time_24h <- as.POSIXct(paste(heart_rate$Time, heart_rate$AM_or_PM), format = '%H:%M:%S')
# Group according to new column Time_24h
heart_rate_avg <- heart_rate %>%
group_by(Time_24h) %>%
summarise(heartrate = mean(Value))
# About 84,000 points remained after summarizing. Close to the 86,400 seconds in 1 day
glimpse(heart_rate_avg)## Rows: 86,046
## Columns: 2
## $ Time_24h <dttm> 2024-06-05 00:00:00, 2024-06-05 00:00:01, 2024-06-05 00:00:…
## $ heartrate <dbl> 65.23214, 64.00000, 63.80000, 65.50000, 75.00000, 65.55769, …# Get current date
current_date <- Sys.Date()
# Highlighted areas on 24h graph
highlight1 <- data.frame(xmin = as.POSIXct(paste(current_date, '12:00:00')),
xmax = as.POSIXct(paste(current_date, '13:30:00')),
ymin = -Inf, ymax = Inf)
highlight2 <- data.frame(xmin = as.POSIXct(paste(current_date, '16:30:00')),
xmax = as.POSIXct(paste(current_date, '19:00:00')),
ymin = -Inf, ymax = Inf)
# Create new table for 1st peak of heart rate
heart_rate_peak1 <- heart_rate %>%
filter(Time_24h >= as.POSIXct(paste(current_date, '12:00:00')) &
Time_24h <= as.POSIXct(paste(current_date, '13:30:00')))
# Create new table for 2nd peak of heart rate
heart_rate_peak2 <- heart_rate %>%
filter(Time_24h >= as.POSIXct(paste(current_date, '16:30:00')) &
Time_24h <= as.POSIXct(paste(current_date, '19:00:00')))# Plot graph of 24h data
ggplot(heart_rate_avg,aes(x = Time_24h,y = heartrate)) +
geom_rect(data = highlight1, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'lightcoral', alpha = 0.5, inherit.aes = FALSE) +
geom_rect(data = highlight2, aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax),
fill = 'lightcoral', alpha = 0.5, inherit.aes = FALSE) +
geom_point(alpha = 0.1) +
geom_smooth(aes(x = Time_24h, y = heartrate), method = "loess", span = 0.1, se=FALSE) +
theme(axis.text.x = element_blank(), axis.title.x = element_text(margin = margin(t = 30))) +
annotate('text', x = as.POSIXct(paste(current_date, '00:00:00')), y = -15, label = '00:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '06:00:00')), y = -15, label = '06:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:00:00')), y = -15, label = '12:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '18:00:00')), y = -15, label = '18:00h') +
annotate('text', x = as.POSIXct(paste(current_date, '23:59:00')), y = -15, label = '23:59h') +
annotate('text', x = as.POSIXct(paste(current_date, '12:30:00')), y = 50, label = '(12:00-13:30h)',
colour = 'darkred') +
annotate('text', x = as.POSIXct(paste(current_date, '17:45:00')), y = 50, label = '(16:30-19:00h)',
colour = 'darkred') +
coord_cartesian(ylim = c(0, 150), clip = 'off') +
labs(x = 'Time of Day', y = 'Heart rate', title = 'Hourly Average Heart Rate') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5)) The heart rate spikes up between 12:00h-13:30h and 16:30h-19:00h, consistent with the increased steps and intensity at similar times as discussed in section 4.5 and section 4.6, respectively.
Key Findings
- Heart rate peaks between 12pm-1:30pm and 4:30pm-7pm
4.10.2 Analyzing the 1st peak heart rate
I created a table to view the heart rate at the 1st peak around 12:00h
# Summarize 1st peak segment of 24h data
heart_rate_peak1_avg <- heart_rate_peak1 %>%
group_by(Time_24h) %>%
summarise(heartrate = mean(Value))# Plot graph of 1st peak
ggplot(heart_rate_peak1_avg,aes(x = Time_24h,y = heartrate)) +
geom_point(alpha = 0.5) +
geom_smooth(method = "loess", se = FALSE, span = 0.5, linewidth = 2) +
labs(x = 'Time of Day', y = 'Heart rate', title = 'Peak Heart Rate at 12:30h') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5))A huge portion of FitBit users had an increase in heart rate around noon.
Key Findings
- Heart rate reaches an average of 100 beats per minute at 12:30pm daily for FitBit users
4.10.3 Analyzing the 2nd peak heart rate
I created a table to view the heart rate at the 2nd peak around 17:00h and 18:30h.
# Summarize 2nd peak segment of 24h data
heart_rate_peak2_avg <- heart_rate_peak2 %>%
group_by(Time_24h) %>%
summarise(heartrate = mean(Value))# Plot graph of 2nd peak
ggplot(heart_rate_peak2_avg,aes(x = Time_24h,y = heartrate)) +
geom_point(alpha = 0.1) +
geom_smooth(method = 'loess', se = FALSE, span = 0.2, linewidth = 2) +
labs(x = 'Time of Day', y = 'Heart rate', title = 'Peak Heart Rates at 17:00h and 18:30h') +
theme(axis.title.x = element_text(size = 15),
axis.title.y = element_text(size = 15),
plot.title = element_text(size = 15, hjust = 0.5)) A large amount of FitBit users also had higher heart rates around 17:00h and 18:30h
Key Findings
- The heart rate also reaches an average of about 100 beats per minute both at 4:45pm and 6:30pm daily
Phase 5: Act (Conclusion)
According to my analysis, I propose three marketing strategy recommendations for Bellabeat’s Time watch, each substantiated by data and accompanied by actionable steps.
1. Appeal to the broader population by highlighting the sleep tracking capabilities of the Time watch
| Data | FitBit users typically start going to bed by 8:00 pm, with most users waking up starting at 05:30 am. Additionally, the average sleep duration is approximately 7 hours per day, with users getting the most sleep on Sundays. |
| Actionable Steps |
|
| Limitations | Potential users may be turned off by the high price of Time watches since the device may come with other features they are not particularly interested in |
2. Match the timing of marketing strategies to coincide with the average schedules and peak activity times of users based on the day of the week
| Data | FitBit user data shows consistent usage of the device throughout the week, with peaks in calorie burn, intensity, and heart rate occurring around 12:00 pm and 7:00 pm. Steps taken and calories burned tend to be the lowest on Sundays and the highest on Saturdays, indicating potential opportunities to encourage wellness habits strategically. |
| Actionable Steps |
|
| Limitations |
|
3. Promote the consistent usage of the Time watch to improve fitness results:
| Data | Those who consistently wore their FitBit device even during sleep demonstrated a sustained increase in physical activity over time compared to those who didn’t record their sleep. |
| Actionable Steps |
|
| Limitations | From the subgroup of FitBit users mentioned in section 4.8.2: Their increase in light physical activity over time may not actually be due to them wearing the FitBit device, but rather due to their inherent motivation to be healthier. This motivation may have simply just affected their choice to record every aspect of their journey, including their sleep patterns (correlation, not causation). |
By incorporating these recommendations into its marketing strategy, Bellabeat can effectively leverage the insights from FitBit user data to position the Time watch as a comprehensive solution for tracking and optimizing users’ health, fitness, and sleep routines throughout the week.