PA1_template.Rmd

---
title: "Reproducible Research: Peer Assessment 1"
output: 
  html_document:
    keep_md: true
---


## Loading and preprocessing the data

The data is unzipped, loaded, and reshaped using the melt function in the reshape2 package.

```{r, echo=TRUE}
library(reshape2)
unzip("activity.zip")
activity <- read.csv("activity.csv")
act <- melt(activity, id.vars=c("date", "interval"))
head(activity)
head(act)
```

## What is mean total number of steps taken per day?

Ignoring missing values, a histogram of total number of steps taken per day is plotted:

```{r hist1, echo=TRUE}
actsum <- dcast(act, date ~ variable, fun.aggregate = sum, na.rm=TRUE)
head(actsum)
with(actsum, hist(steps, breaks=20))
```

The mean and median values of the total number of steps taken per day is calculated:

```{r, echo=TRUE}
mean(actsum$steps)
median(actsum$steps)
```

## What is the average daily activity pattern?

Still ignoring missing values, the average number of steps taken in a five-minute interval (averaged over all days) is plotted as a time series:

```{r avsteps1, echo=TRUE}
actint <- dcast(act, interval ~ variable, fun.aggregate = mean, na.rm=TRUE)
head(actint)
with(data=actint, plot(x=interval, y=steps, type="l", main="Average steps per interval"))
```

The 5-minute interval with the maximum number of steps, averaged across all days, is found here:

```{r, echo=TRUE}
actint[which.max(actint$steps),]
```

## Imputing missing values

The total number of missing steps values is calculated:

```{r, echo=TRUE}
sum(is.na(act$value))
```

To impute the missing number of steps for a given time interval, the average value of steps for that time interval (averaged over all days) is used:

```{r, echo=TRUE}
filled <- act
for (i in 1:length(filled$value)) {
  if (is.na(filled[i,]$value)) {
    int = filled[i,]$interval
    filled[i,]$value <- actint[actint$interval==int,]$steps
  }
}
actsum_filled <- dcast(filled, date ~ variable, fun.aggregate = sum, na.rm=TRUE)
head(actsum_filled)
```

The histogram for total number steps per day (with missing values imputed) is plotted and the mean and median values are calculated:

```{r hist2, echo=TRUE}
with(actsum_filled, hist(steps, breaks=20, main="Histogram of steps (with imputed missing values)"))
mean(actsum_filled$steps)
median(actsum_filled$steps)
```

Imputing values decreases the frequency of 0 values and increases the frequency of values at the peak around 10,000, as seen from a comparison of histograms.  This increases the values for mean and median as well.

## Are there differences in activity patterns between weekdays and weekends?

A new factor variable is created that indicates whether a given date is a weekday or weekend day, and the data is reshaped:

```{r, echo=TRUE}
filled$day <- ifelse(weekdays(as.Date(filled$date)) %in% c("Saturday", "Sunday"),
                     "weekend", "weekday")
weekdaysteps <- dcast(filled, interval ~ day, fun.aggregate = mean)
head(weekdaysteps)
weekdaymelt <- melt(weekdaysteps, id.vars = "interval")
head(weekdaymelt)
```

Finally, the average number of steps taken during a given 5-minute interval, averaged over all weekend days (top) or over all weekdays (bottom) are plotted as time series plots:

```{r comparison, echo=TRUE}
library(lattice)
xyplot(value~interval | variable, weekdaymelt, type="l", layout=c(1,2), xlab="Interval", ylab="Number of Steps", main="Average steps per interval on weekdays or weekend days")
```

The plots demonstrate that there are different patterns on weekdays and weekends, on average.