Skip to content

Latest commit

 

History

History
208 lines (149 loc) · 4.88 KB

PA1_template.md

File metadata and controls

208 lines (149 loc) · 4.88 KB

Reproducible Research: Peer Assessment 1

Erik Konijnenburg

Loading and preprocessing the data

First we download the zipfile from the website if it does not exist on the disk. Then we unzip the file. Then we read in the csvDataFile. As a last step we convert the date column to a Date type.

zipfile<-"./activity.zip"
url<-"https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2Factivity.zip"
if(!file.exists(zipfile)) {
    download.file(url, zipfile, method="libcurl")
}

unzip(zipfile)
csvDataFile<-"./activity.csv"
activity<-read.csv(csvDataFile, header = TRUE, sep=",")

library(lubridate)
## 
## Attaching package: 'lubridate'

## The following object is masked from 'package:base':
## 
##     date

activity$date<-ymd(activity$date)

What is mean total number of steps taken per day?

Next the total number of steps per day is determined. From this result the mean and median number of steps is calculated.

library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:lubridate':
## 
##     intersect, setdiff, union

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

stepsPerDay <- activity %>%
               group_by(date) %>% 
               summarize(totalsteps = sum(steps, na.rm=TRUE))


hist(x=stepsPerDay$totalsteps, breaks = 10, xlab="Total steps per day. ", 
        main = "Histogram of total steps per day")

meanStepsPerDay <- mean(stepsPerDay$totalsteps)
meanStepsPerDay
## [1] 9354.23

medianStepsPerDay <- median(stepsPerDay$totalsteps)
medianStepsPerDay
## [1] 10395

What is the average daily activity pattern?

To determine the average number of steps per interval the data is grouped by interval and the mean of all steps in each group is taken.

stepsPerInterval <- activity %>%
               group_by(interval) %>% 
               summarize(averageSteps = mean(steps, na.rm=TRUE))

plot(stepsPerInterval$interval,stepsPerInterval$averageSteps,type="l", 
            xlab="Interval", ylab="Number of steps", main="Average number of steps per interval")

Next the interval with the maximum number of steps is determined

maxIndex <- stepsPerInterval$averageSteps==max(stepsPerInterval$averageSteps)
stepsPerInterval[maxIndex,]$interval
## [1] 835

Imputing missing values

The initial dataset contains NA values. Let's calculate the amount of NA values

naIndex <- is.na(activity$steps)
sum(naIndex)
## [1] 2304

For imputing the missing values a simple strategy is choosen. The strategy is to take the average step value of that interval (averaged over all days) if the value is missing.

funcImputeValue <- function(x) {
    stepsPerInterval[stepsPerInterval["interval"] == x,]$averageSteps
}

imputedActivities <- activity %>% 
    mutate(steps = ifelse(
        is.na(steps), funcImputeValue(interval), steps))

stepsPerDayWithImpute <- imputedActivities %>%
               group_by(date) %>% 
               summarize(totalsteps = sum(steps))


hist(x=stepsPerDayWithImpute$totalsteps, breaks = 10, xlab="Total steps per day (with imputed NA values). ", 
        main = "Histogram of total steps per day")

meanStepsPerDayImputed <- mean(stepsPerDayWithImpute$totalsteps)
meanStepsPerDayImputed
## [1] 10766.19

medianStepsPerDayImputed <- median(stepsPerDayWithImpute$totalsteps)
medianStepsPerDayImputed
## [1] 10766.19

By imputing values the both median and mean values change compared to the initial calculation of these values without imputing missing values. Imputing results in a median that is closer to the average.

Are there differences in activity patterns between weekdays and weekends?

First a weekend/weekday label is assigned to the imputed dataset. Then the average of the steps per interval and weekday is calculated.

imputedActivities <- imputedActivities %>%
        mutate(weekday=
                as.factor(ifelse(weekdays(date) == "Saturday" | weekdays(date) == "Sunday", 
                "weekend", "weekday" ) ))

stepsPerIntervalWithWeekday <- imputedActivities %>%
               group_by(interval, weekday) %>% 
               summarize(meanstep = mean(steps))

Finally the steps as function of the interval is plotted as for both weekdays and weekend

library(lattice)
xyplot(meanstep ~ interval | factor(weekday), 
       data=stepsPerIntervalWithWeekday, type="l", layout=c(1,2), xlab = "Number of steps", 
       ylab="Steps" )