Author: Daniel Kinpara
Date: April, 2015

1. Loading and preprocessing the data

The first step was to read the file activity.zip and to save it in the object dados. The ZIP file must be placed in the same folder of this Markdown document in order to run the embedded code. The variable date was converted to a date class.

if (!file.exists("activity.csv")) {
        unzip("activity.zip")
}
dados <- read.csv("activity.csv", header = TRUE, na.strings = "NA")
dados[,2] <- as.Date(dados[,2], "%Y-%m-%d")
str(dados)
## '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 ...

2. What is mean total number of steps taken per day?

The code below created a factor variable named dia that was used to split the dataset. The tapply summed the number of steps per day. The average was calculated and stored in the variable mediaPassoDia. The median was stored in the variable medianaPassoDia.

dados <- transform(dados, dia = as.factor(dados$date))
totalPassoDia <- tapply(dados$steps, dados$dia, sum, na.rm = TRUE)
mediaPassoDia <- mean(totalPassoDia)
medianaPassoDia <- median(totalPassoDia)
y <- as.integer(totalPassoDia)
x <- as.Date(names(totalPassoDia))
plot(x, y, type="h", xlab="Day", ylab="Total steps per day")

So, the average number of steps per day was 9354. The median was 10395. The missing data was excluded from the calculation.

3. What is the average daily activity pattern?

The code created a factor variable named tempo that was used to split the dataset. The tapply averaged the number of steps per interval for all measured days. The average was stored in the variable mediaPassoTempo.

dados <- transform(dados, tempo = as.factor(dados$interval))
mediaPassoTempo <- tapply(dados$steps, dados$tempo, mean, na.rm = TRUE)
y <- as.integer(mediaPassoTempo)
x <- names(mediaPassoTempo)
plot(x, y, type = "l", xlab = "Interval", ylab = "Average steps per 5 min interval")

The highest average steps occurred in the 835 interval. The missing data was excluded from the average calculation.

Imputing missing values

The missing value of a giving interval was substituted by the average of the same interval across the measured days. The following code created a new dataset called dadosSemNA. It checks wether the row has a NA in the steps variable. If it does, the respective interval average is searched in the variable mediaPassoTempo.

numMedidas <- length(dados$steps)
dadosSemNA <- dados
for(i in 1:numMedidas){
        if(is.na(dadosSemNA[i, 1])){
                dadosSemNA[i,1] <- mediaPassoTempo[as.character(dadosSemNA[i, 5])]
        }
}
totalPassoDia <- tapply(dadosSemNA$steps, dadosSemNA$dia, sum)
mediaPassoDia <- mean(totalPassoDia)
medianaPassoDia <- median(totalPassoDia)
y <- as.integer(totalPassoDia)
x <- as.Date(names(totalPassoDia))
plot(x, y, type="h", xlab="Day", ylab="Total steps per day")

After missing data replacement, the average number of steps per day was 10766. The median was 10766. Both values were higher than the ones obtained with the dataset with missing values.

Are there differences in activity patterns between weekdays and weekends?

The code below created a new column called semana which identifies the weekdays (level 1) and weekends (level 2). Through the function aggregate was calculated the mean for each group, weekday and weekend.

dadosSemNA <- transform(dadosSemNA, semana = factor(rep(1, numMedidas), levels = c(1, 2), labels = c("weekday", "weekend")))
for(i in 1:numMedidas){
        if(weekdays(dadosSemNA[i, 2]) == "Sunday" | weekdays(dadosSemNA[i, 2]) == "Saturday"){
                dadosSemNA[i, 6] <- "weekend"
        }
        else{
                dadosSemNA[i, 6] <- "weekday"
        }
}
dadosSplit <- with(dadosSemNA, aggregate(x = steps, by = list(semana, tempo), FUN = "mean"))
par(mfrow = c(2,1))
x <- levels(dadosSplit$Group.2)
y1 <- dadosSplit[dadosSplit$Group.1 == "weekday", 3] 
y2 <- dadosSplit[dadosSplit$Group.1 == "weekend", 3]
mediaY1 <- mean(y1)
mediaY2 <- mean(y2)
plot(x, y1, type = "l", main = "Weekdays", xlab = "Interval", ylab = "Average steps per 5 min interval", ylim = c(0, 250))
abline(mediaY1, 0, col = "red")
plot(x, y2, type = "l", main = "Weekends", xlab = "Interval", ylab = "Average steps per 5 min interval", ylim = c(0, 250))
abline(mediaY2, 0, col = "red")

The results showed that the activity patterns were different. During the weekdays, the individual took more steps in a short period of time, concentrated in the begining of the day and in the end of the day. However, in the weekends, one took less steps per period of time, but constantly along the day. So, the averages along the day were not so different. During the weekdays, the subject took an average of 36 steps per 5 min interval along the day. During the weekends, the average was 42 steps. The total steps per day taken during the weekdays was 10256 steps and, during the weekends, it was 12202 steps. So, the subject walked per day more during the weekends.