Activity Monitoring Project

Loading and Processing Data

Use read.csv to load the data into a variable, called activity. Print the column names of the data to see what you’re working with.

library(dplyr)
library(tidyr)
library(chron)
library(lattice)
library(ggplot2)
activity <- read.csv("activity.csv")
activity <- tbl_df(activity)
colnames(activity)

## [1] "steps"    "date"     "interval"

Summarizing the steps data

Make a histogram of the steps column of activity to visualize the data.

hist(activity$steps, xlab="Steps")

Find the mean and median of the steps data.

Use the mean and median functions on the steps column of the activity data.

step_mean <- mean(activity$steps, na.rm=TRUE)
step_med <- median(activity$steps, na.rm=TRUE)
{print(step_mean)
 print(step_med)}

## [1] 37.3826
## [1] 0

Mean of steps data: 37.3825996

Median of steps data: 0

Visualizing time intervals over days

Group the data by interval. Use the summarize function to group the activity data, and find the mean of each interval.

time_interval_means <- summarize(group_by(activity, interval), mean(steps, na.rm=TRUE))

Use the base plotting system to plot the average steps by time interval.

plot(time_interval_means, type="l",xlab="Time Interval", ylab = "Average Steps")

To find the time interval with the highest average steps, use the arrange function to arrange the time_interval_means by descending average steps. The first row will be the time interval with the highest average steps.

colnames(time_interval_means) <- c("interval", "steps")
time_interval_means <- arrange(time_interval_means, desc(steps))

Interval with highest average steps: 835

(206.1698113 steps)

Imputing Missing Values

To impute missing values, I made each NA steps value equal to the average steps value for that time interval.

time_interval_means <- arrange(time_interval_means, interval)
activity_imputed <- activity
for(i in 1:nrow(activity)){
  if(is.na(activity[i,1])){
    activity_imputed[i,1] <- as.numeric(time_interval_means[(activity[i,3]/5)+1,2])
  }
}
hist(activity_imputed$steps, xlab="Steps")

step_mean_imputed <- mean(activity_imputed$steps, na.rm=TRUE)
step_med_imputed <- median(activity_imputed$steps, na.rm=TRUE)
{print(step_mean_imputed)
 print(step_med_imputed)}

## [1] 37.6206
## [1] 0

Mean of steps data with imputed missing values: 37.6206044

Median of steps data with imputed missing values: 0

Looking at Weekdays vs. Weekends

Use dplyr’s mutate function to split the date column into two factors, TRUE for days that are weekends, or FALSE for days that are during the work week. Store this in a new column called day. Then, use the factor function to relabel day with Weekday and Weekend. Use lattice’s xyplot function to create a time plot of the average steps per interval split over Weekends and Weekdays

dayActivity <- mutate(activity, day=is.weekend(as.Date(activity$date)))
dayActivity$day <- factor(dayActivity$day, label=c("Weekday", "Weekend"))
dayActivity <- summarize(group_by(dayActivity, interval, day), mean(steps, na.rm=TRUE))
colnames(dayActivity) <- c("interval", "day", "steps")
xyplot(steps ~ interval | day, col="#68ab98", type="l", layout=c(1,2), data=dayActivity)

The the steps data for weekends and for weekdays are clearly different. On weekends, the steps taken between the \(500^{th}\) and \(2000^{th}\) time intervals are consistently higher than the other time intervals. On weekdays, there is a sizeable spike from the \(500^{th}\) to the \(1000^{th}\) time interval.