Background

Fitbit is an American consumer electronics and fitness company. It produces wireless-enabled wearable technology physical fitness monitors and activity trackers such as smartwatches, pedometers and monitors for heart rate, quality of sleep, and stairs climbed as well as related software.

Data Source

This Kaggle dataset compiles many different activity/sleep/calories stats from users in 2016. I analyzed this dataset with R programming and used RStudio to build custom visualizations to understand correlations and trends with the data.

Setting Up The Environment

Installation of R packages needed to perform extraction & analysis

install.packages("tidyverse")
library(tidyverse)
library(readr)
library(lubridate)
install.packages("skimr")
library(skimr)
library(dplyr)

==================================================
downloaded 413 KB
* installing *binary* package ‘tidyverse’ ...
* DONE (tidyverse)

The downloaded source packages are in
    ‘/tmp/RtmpvIVAeB/downloaded_packages’
    
> library(tidyverse)
── Attaching packages ──────────────────────────────────────────────────────────────────── tidyverse 1.3.1 ──
✓ ggplot2 3.3.3     ✓ purrr   0.3.4
✓ tibble  3.1.0     ✓ dplyr   1.0.5
✓ tidyr   1.1.3     ✓ stringr 1.4.0
✓ readr   1.4.0     ✓ forcats 0.5.1
── Conflicts ─────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
x dplyr::filter() masks stats::filter()
x dplyr::lag()    masks stats::lag()

Loading & Previewing the Raw Data

daily_activity <- read_csv("dailyActivity_merged.csv")
View(daily_activity)
glimpse(daily_activity)
colnames(daily_activity)

daily_activity
 [1] "Id"                       "Date"                     "TotalSteps"              
 [4] "TotalDistance"            "TrackerDistance"          "LoggedDistance"          
 [7] "VeryActiveDistance"       "ModeratelyActiveDistance" "LightActiveDistance"     
[10] "SedentaryActiveDistance"  "VeryActiveMinutes"        "FairlyActiveMinutes"     
[13] "LightlyActiveMinutes"     "SedentaryMinutes"         "Calories"                
[16] "WeekDay"

daily_steps <- read_csv("dailySteps_merged.csv")
View(daily_steps)
head(daily_steps)

A tibble: 6 x 3
          Id ActivityDay StepTotal
       <dbl> <chr>           <dbl>
1 1503960366 4/12/2016       13162
2 1503960366 4/13/2016       10735
3 1503960366 4/14/2016       10460
4 1503960366 4/15/2016        9762
5 1503960366 4/16/2016       12669
6 1503960366 4/17/2016        9705

daily_calories <- read_csv("dailyCalories_merged.csv")
View(daily_calories)
skim_without_charts(daily_calories)

── Data Summary ────────────────────────
                           Values        
Name                       daily_calories
Number of rows             940           
Number of columns          3             
_______________________                  
Column type frequency:                   
  character                1             
  numeric                  2             
________________________                 
Group variables            None          
...

daily_sleep <- read_csv("sleepDay_merged.csv")
View(daily_sleep)
summary(daily_sleep)

       Id              SleepDay         TotalSleepRecords TotalMinutesAsleep TotalTimeInBed 
 Min.   :1.504e+09   Length:413         Min.   :1.000     Min.   : 58.0      Min.   : 61.0  
 1st Qu.:3.977e+09   Class :character   1st Qu.:1.000     1st Qu.:361.0      1st Qu.:403.0  
 Median :4.703e+09   Mode  :character   Median :1.000     Median :433.0      Median :463.0  
 Mean   :5.001e+09                      Mean   :1.119     Mean   :419.5      Mean   :458.6  
 3rd Qu.:6.962e+09                      3rd Qu.:1.000     3rd Qu.:490.0      3rd Qu.:526.0  
 Max.   :8.792e+09                      Max.   :3.000     Max.   :796.0      Max.   :961.0

Understanding our Data

How many unique participants are in the study?

n_distinct(daily_activity$Id)
n_distinct(daily_sleep$Id)
n_distinct(daily_steps$Id)
n_distinct(daily_calories$Id)

[1] 33 [1] 24 [1] 33 [1] 33 - The daily sleep study has the least amount of participants

How many observations are there in each dataframe?

nrow(daily_activity)
nrow(daily_sleep)
nrow(daily_steps)
nrow(daily_calories)

[1] 940 [1] 410 [1] 940 [1] 940 - The daily sleep study has the least amount of observations

How many unique ID’s are across all dataframes?

length(unique(do.call(c, list(daily_activity$Id, daily_steps$Id, daily_sleep$Id, daily_calories$Id))))

[1] 33 - There are 33 unique ID’s for participants

Data Cleaning

Creating consistent column names

daily_sleep %>% 
  select(SleepDay)
help(lubridate)

Rename(dataframe, new name = old name)

Converting datetime to only date

daily_sleep <- dplyr::rename(daily_sleep, SleepDate = SleepDay) %>%
  daily_sleep$SleepDate <- as.Date(daily_sleep$SleepDate, "%d%m%Y")

Checking to see if the time was successfully removed

daily_sleep %>% 
  select(SleepDate)
daily_sleep$SleepDate <- as.Date(daily_sleep$SleepDate, "%d%m%Y")
daily_sleep$SleepDate <- strptime(daily_sleep$SleepDate,format = "%d/%m/%Y") 
daily_sleep$SleepDate <- as.Date(daily_sleep$SleepDate = "%d%m%Y")
class(daily_sleep$SleepDate)

Update: this date function did not work because the class of SleepDate column is string rather than date
Need to brainstorm a different solution to removing the time

daily_sleep <- separate(daily_sleep, col=SleepDate, into=c('SleepDate','Time','AMPM'), sep=' ') 
## this separation worked, now we will extract & separate month, day, and year
daily_sleep <- separate(daily_sleep, col=SleepDate, into=c('Month','Day','Year'), sep='/') 
daily_sleep %>% 
  select(Year)

We can delete the time and AM/PM columns since these are always 12AM (irrelevant)

Output:

# A tibble: 6 x 7
          Id   Month Day   Year  TotalSleepRecords TotalMinutesAsleep TotalTimeInBed
       <dbl>   <chr> <chr> <chr>             <dbl>              <dbl>          <dbl>
1 1503960366   4     12    2016                  1                327            346
2 1503960366   4     13    2016                  2                384            407
3 1503960366   4     15    2016                  1                412            442 ...

daily_sleep %>% select(-AMPM, -Time)
daily_sleep <- select(daily_sleep, -AMPM, -Time)

Permanent removal and stored in daily_sleep

Fixing daily_activity columns and extracting weekday from date

daily_activity <-
  rename(daily_activity, Date = ActivityDate, LoggedDistance = LoggedActivitiesDistance) %>%
  mutate(Date = mdy(Date), WeekDay = weekdays(Date))
colnames(daily_activity)
class(daily_activity$Date)

checking to ensure the weekday extraction worked correctly

daily_activity %>% 
  select(WeekDay)

Performing same steps for other columns

daily_steps <-
  rename(daily_steps, Date = ActivityDay, TotalSteps = StepTotal) %>%
  mutate(Date = mdy(Date), WeekDay = weekdays(Date))
daily_steps %>% 
  select(WeekDay)

worked successfully

daily_calories <-
  rename(daily_calories, Date = ActivityDay) %>%
  mutate(Date = mdy(Date), WeekDay = weekdays(Date))
daily_calories %>% 
  select(WeekDay)

Checking for NULLs and Duplicates

> print(paste("NAs: ", sum(is.na(daily_activity)))) # check for NAs
[1] "NAs:  0"
> print(paste("Duplicated rows: ", sum(duplicated(daily_activity)))) # check for Duplicates
[1] "Duplicated rows:  0"
> print(paste("NAs: ", sum(is.na(daily_calories)))) 
[1] "NAs:  0"
> print(paste("Duplicated rows: ", sum(duplicated(daily_calories))))
[1] "Duplicated rows:  0"
> print(paste("NAs: ", sum(is.na(daily_steps)))) 
[1] "NAs:  0"
> print(paste("Duplicated rows: ", sum(duplicated(daily_steps))))
[1] "Duplicated rows:  0"
> print(paste("NAs: ", sum(is.na(daily_sleep))))
[1] "NAs:  0"
> print(paste("Duplicated rows: ", sum(duplicated(daily_sleep))))
[1] "Duplicated rows:  3"

3 Duplicated rows in daily_sleep dataframe, will delete

daily_sleep <- daily_sleep[!duplicated(daily_sleep), ]
print(paste("Duplicated rows: ", sum(duplicated(daily_sleep))))

> print(paste("Duplicated rows: ", sum(duplicated(daily_sleep))))
[1] "Duplicated rows:  0"

Duplicates have been removed

Merging data to find correlations/patterns

Chose Id as the primary key since they all share this in common

daily_activity_sleep <- merge(daily_activity, daily_sleep, by="Id")
daily_steps_calories <- merge(daily_steps, daily_calories, by="Id")
colnames(daily_activity_sleep)
colnames(daily_steps_calories)

> colnames(daily_activity_sleep)
 [1] "Id"                       "Date"                     "TotalSteps"               "TotalDistance"           
 [5] "TrackerDistance"          "LoggedDistance"           "VeryActiveDistance"       "ModeratelyActiveDistance"
 [9] "LightActiveDistance"      "SedentaryActiveDistance"  "VeryActiveMinutes"        "FairlyActiveMinutes"     
[13] "LightlyActiveMinutes"     "SedentaryMinutes"         "Calories"                 "WeekDay"                 
[17] "Month"                    "Day"                      "Year"                     "TotalSleepRecords"       
[21] "TotalMinutesAsleep"       "TotalTimeInBed"    

> colnames(daily_steps_calories)
[1] "Id"         "Date.x"     "TotalSteps" "WeekDay.x"  "Date.y"     "Calories"   "WeekDay.y" 
# Digging into average steps, average calories, and average sleep per day
daily_average <- daily_activity_sleep %>%
  group_by(Id) %>%
  summarise(AvgSteps = mean(TotalSteps), AvgCalories = mean(Calories), AvgSleep = mean(TotalMinutesAsleep))
head(daily_average)

Calculating average activity (steps, distance, sleep) grouped by weekday

weekday_average <- daily_activity_sleep %>%
  group_by(WeekDay) %>%
  summarize(AvgSteps = mean(TotalSteps), AvgDistance = mean(TotalDistance), AvgSleep = mean(TotalMinutesAsleep), AvgInactivity = mean(TotalTimeInBed)) 

weekday_avg_cals <- daily_calories %>%
  group_by(WeekDay) %>%
  summarize(AvgCalories = mean(Calories))

Ordering the week days

weekday_average$WeekDay <- c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday")
weekday_avg_cals$WeekDay <- c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday")

Proof below to ensure the weekday extraction and averages worked successfully

> head(weekday_average)
# A tibble: 6 x 5
WeekDay   AvgSteps AvgDistance AvgSleep AvgInactivity
<chr>        <dbl>       <dbl>    <dbl>         <dbl>
1 Monday       8231.        5.77     419.          458.
2 Tuesday      8644.        6.10     419.          459.
3 Wednesday    8622.        6.07     419.          458.
4 Thursday     6585.        4.70     419.          459.
5 Friday       7736.        5.46     420.          458.
6 Saturday     9017.        6.34     419.          458.

Data Correlations & Visualizations

Plotting the correlation between Average Daily Sleep and Average Calories Burned shown in a scatter plot

ggplot(data=daily_average, aes(x=AvgCalories, y=AvgSleep)) + 
  geom_point() +
  labs(y="Average Calories", x="Average Sleep", title="Daily Average Sleep Vs. Calories Among Users")

It is not easy here to see the relationship between these two variables, there seems to be no correlation
I will retry with different metrics and a more intuitive scatter plot design

Relationship between Total Sleep Time and Total Time Sedentary (inactive in bed)

ggplot(data=daily_sleep, aes(TotalMinutesAsleep, TotalTimeInBed, fill="origin")) + 
  geom_point(alpha=0.5, size=2, shape=21) +
  labs(y="Minutes Inactive", x="Minutes Asleep", title="Daily Sleep Vs. Inactivity Among Users")

- Strong positive correlation shown here - total time spent in bed and total time asleep go hand-in-hand

Correlation between Total Daily Distance and Daily Calories

ggplot(data=daily_activity, aes(x=TotalDistance, y=Calories)) + 
  geom_point(alpha=0.5, size=2) +
  labs(y="Calories", x="Distance", title="Daily Distance Vs. Calories Burned Among Users")

- Here, we can see a weaker positive correlation between distance and calories burned, minus the outliers

Adding a Trend Line for Clarity

ggplot(data=daily_activity, aes(x=TotalSteps, y=Calories)) + 
  geom_jitter() +
  geom_smooth(color = "green") + 
  labs(y="Calories", x="Steps", title="Daily Steps Vs. Calories Burned Among Users")

- The same is true for the relationship between daily steps and calories

Bar Graphs for Average Activity Stats

Average Steps per weekday

barplot(weekday_average$AvgSteps, main="Weekly Average Steps", horiz = TRUE,
        names.arg=c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"),
        col=c("dark green"),
        density=c(4,5,9,8,7,30,7), angle=c(7,45,90,11,36,6,8))

- Saturday, which I outlined in the textured bar graphs, is the day out of the week with the most step activity

Average Calories per weekday

barplot(weekday_avg_cals$AvgCalories, main="Weekly Average Calories", horiz = TRUE,
        names.arg=c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"),
        col=c("blue"),
        density=c(4,5,9,8,7,30,7), angle=c(7,45,90,11,36,6,8))

- The most two active days for burning calories are Saturday and Wednesday, with Saturday ahead by only 1.045 cals

Average Steps per weekday

barplot(weekday_average$AvgDistance, main="Weekly Average Distance",
        names.arg=c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"),
        col=c("white","light blue","pink","purple","maroon","dark green","black"))

- Saturday, shown in green, is the week day with the most distance covered, which aligns with the steps activity observation

What is the Percentage of Active FitBit Users?

Creating a dataframe to store user steps activity grouped into realistic activity categories

user_activity <- daily_steps %>%
  mutate(UserType = case_when(
    TotalSteps < 5000 ~ "sedentary",
    TotalSteps >= 5000 & TotalSteps < 7499 ~ "lightly active", 
    TotalSteps >= 7500 & TotalSteps < 9999 ~ "fairly active", 
    TotalSteps >= 10000 ~ "very active"))

Creating percentages for each user activity type so it can be later visualized in a pie chart

activity_percentage <- user_activity %>%
  group_by(UserType) %>%
  summarise(total = n()) %>%
  mutate(totals = sum(total)) %>%
  group_by(UserType) %>%
  summarise(TotalPercent = total / totals) %>%
  mutate(Percentages = scales::percent(TotalPercent))

head(activity_percentage)
# A tibble: 4 x 3
  UserType       TotalPercent Percentages
  <chr>                 <dbl> <chr>      
1 very active           0.173 17.34%     
2 fairly active         0.182 18.19%     
3 lightly active        0.322 32.23%     
4 sedentary             0.322 32.23%

Ordering the levels of step activity

activity_percentage$UserType <- c("very active", "fairly active", "lightly active", "sedentary")
head(activity_percentage)

Creating a Pie Chart Visual

activity_percentage %>%
  ggplot(aes(x="",y=TotalPercent, fill=UserType)) +
  geom_bar(stat = "identity", width = 1) +
  coord_polar("y", start=-5, direction=-2) +
  theme_void() +
  theme(plot.title = element_text(hjust = 0.5, face="bold", size=15)) +
  scale_fill_manual(values = c("pink2","yellow2","orange1", "red2")) +
  geom_text(aes(label = Percentages),
            position = position_stack(vjust = 0.5)) +
  ggtitle("Distribution of User Activity")

Noteworthy Observations:

- Here we can see that most users of the fitness app are either sedentary(inactive) or lightly active
- Fairly active and very active users are the minority, only making up 35.53% of the overall 100%
- Saturday is the most active day of the week for users
- There are positive correlations between total time spent in bed & total time asleep, calories & distance, steps & calories 
- Sleep statistics are very similar across the span of a week, ranging from 418-420 minutes for sleep averages

FitBit Study - Visualization with R

Isabel Tummino

11/10/2021

Background

Data Source

Setting Up The Environment

Installation of R packages needed to perform extraction & analysis

Loading & Previewing the Raw Data

Understanding our Data

How many unique participants are in the study?

How many observations are there in each dataframe?

How many unique ID’s are across all dataframes?

Data Cleaning

Creating consistent column names

Rename(dataframe, new name = old name)

Converting datetime to only date

Output:

Fixing daily_activity columns and extracting weekday from date

checking to ensure the weekday extraction worked correctly

Performing same steps for other columns

Checking for NULLs and Duplicates

Merging data to find correlations/patterns

Calculating average activity (steps, distance, sleep) grouped by weekday

Ordering the week days

Data Correlations & Visualizations

Plotting the correlation between Average Daily Sleep and Average Calories Burned shown in a scatter plot

Relationship between Total Sleep Time and Total Time Sedentary (inactive in bed)

Correlation between Total Daily Distance and Daily Calories

Adding a Trend Line for Clarity

Bar Graphs for Average Activity Stats

Average Steps per weekday

Average Calories per weekday

Average Steps per weekday

What is the Percentage of Active FitBit Users?

Creating a dataframe to store user steps activity grouped into realistic activity categories

Creating percentages for each user activity type so it can be later visualized in a pie chart

Ordering the levels of step activity

Creating a Pie Chart Visual

Noteworthy Observations: