PA1.md

Reproducible Research: Peer Assessment 1

github.com/l0neJT

14 June 2014

Introduction

This assignment demonstrates the use of R Markdown in conjunction with the knitr package to produce literate code. Forked from rdpeng/RepData_PeerAssessment1.

Loading and preprocessing the data

1. Load packages plyr and ggplot2

   library(plyr)
   library(ggplot2)

2. Load data from file activity.csv

   dat <- read.csv("./data/activity.csv")

3. Transform the 'date' column from factor into date

   dat <- transform(dat, date = as.Date(date, format = "%Y-%m-%d"))
   str(dat)

   ## 'data.frame':	17568 obs. of  3 variables:
   ##  $ steps   : int  NA NA NA NA NA NA NA NA NA NA ...
   ##  $ date    : Date, format: "2012-10-01" "2012-10-01" ...
   ##  $ interval: int  0 5 10 15 20 25 30 35 40 45 ...
   

4. Remove NAs for first analyses

   datNoNA <- dat[!is.na(dat$steps), ]

What is mean total number of steps taken per day?

1. Summarise steps by date (excluding NAs)

   stepsDate <- ddply(datNoNA, "date", summarise, steps = sum(steps))

2. Chart steps by date as a bar plot with mean and median lines

   histStepsDate <- function(data) {
       # Bar plot of steps by date without NAs
       p <- qplot(date, steps, data = data, geom = "bar", stat = "identity")
       
       # Horizontal line at mean
       pMean <- mean(data$steps)
       p <- p + geom_hline(yintercept = pMean, color = "red", size = 2)
       # Label mean line
       posX <- with(data, date[ceiling(length(date) / 2)])
       posY <- 20000
       label <- paste("Mean =", format(pMean, nsmall = 1))
       p <- p + annotate("text", x = posX, y = posY, label = label,
                         color = "red")
       
       # Horizontal line at median
       pMedian <- median(data$steps)
       p <- p + geom_hline(yintercept = pMedian, color = "blue", linetype = 2,
                           size = 1)
       # Label median line
       posY <- 19000
       label <- paste("Median =", format(pMedian, nsmall = 1))
       p <- p + annotate("text", x = posX, y = posY, label = label,
                         color = "blue")
       
       # Return plot
       p
   }
   histStepsDate(stepsDate) + labs(title = "Total Steps by Date (Remove NAs)")

   

What is the average daily activity pattern?

1. Summarise steps by interval (excluding NAs)

   stepsInterval <- ddply(datNoNA, "interval", summarise, steps = mean(steps))

2. Chart steps by interval as a line with max

   lineStepsInterval <- function(data) {
       # Bar plot of steps by date without NAs
       p <- qplot(interval, steps, data = data, geom = "line")
       
       # Label max
       stepsMax <- max(data$steps)
       intvMax <- subset(data, steps == stepsMax, select = interval)[[1]]
       posX <- intvMax + 825L
       posY <- stepsMax
       label <- paste("Interval =", intvMax, "| Max Avg Steps = ",
                      format(stepsMax))
       p <- p + annotate("point", x = intvMax, y = stepsMax, color = "red",
                         size = 5)
       p <- p + annotate("text", x = posX, y = posY, color = "red",
                         label = label)
       
       # Return plot
       p
   }
   lineStepsInterval(stepsInterval) + labs(title =
                                           "Mean Steps by Interval (Remove NAs)")

   

Imputing missing values

1. Replace NAs with mean for interval

   replaceNA <- function(sSteps, iInterval, data = stepsInterval) {
       ifelse(is.na(sSteps),
              subset(data, iInterval == interval, select = steps)[[1]],
              sSteps)
   }
   datReplaceNA <- ddply(dat, c("date", "interval"), transform,
                         steps = replaceNA(steps, interval))

2. Summarize steps by date and chart as a bar plot with mean and median lines

   datTemp <- ddply(datReplaceNA, "date", summarise, steps = sum(steps))
   histStepsDate(datTemp) + labs(title = "Total Steps by Date (Replace NAs)")

   

   rm(datTemp)

3. Summarize steps by interval and chart as a line with max

   datTemp <- ddply(datReplaceNA, "interval", summarise, steps = mean(steps))
   lineStepsInterval(datTemp) + labs(title =
                                     "Mean Steps by Interval (Replace NAs)")

   

   rm(datTemp)

Are there differences in activity patterns between weekdays and weekends?

1. Add weekday character colum to data

   weekDayOrEnd <- function(date) {
       ifelse(format(date, format = "%w") %in% 1:5,
              "weekday",
              "weekend")
   }
   datReplaceNA <- transform(datReplaceNA, weekday = weekDayOrEnd(date))

2. Summarise steps by interval categorized by weekday/weekend

   stepsSplit <- ddply(datReplaceNA, c("interval", "weekday"), summarise,
                       steps = mean(steps))

3. Chart steps by interval as a line with facets by weekday/weekend

   p <- qplot(interval, steps, data = stepsSplit, geom = c("line", "smooth"),
              facets = . ~ weekday)
   p + labs(title = "Mean Steps by Interval (Replace NAs)")

   ## geom_smooth: method="auto" and size of largest group is <1000, so using loess. Use 'method = x' to change the smoothing method.
   ## geom_smooth: method="auto" and size of largest group is <1000, so using loess. Use 'method = x' to change the smoothing method.
   

   

   rm(p)