The data was loaded:
data <- read.csv("./activity.csv")To find the total number of steps per day, the function ddply() from the plyr package is used to summarise the data:
library("plyr")
total.steps <- ddply(data, .(date), summarise, total.steps = sum(steps))For plotting two additional factors are added, one for months and another for days of the week:
total.steps$month <- months(as.POSIXlt(total.steps$date))
total.steps$month <- factor(total.steps$month, levels = c("October", "November"))
total.steps$day <- weekdays(as.POSIXlt(total.steps$date))
total.steps$day <- factor(total.steps$day, levels = c("Monday", "Tuesday", "Wednesday",
"Thursday", "Friday", "Saturday",
"Sunday"))A barplot of the data is created with ggplot, bar colours generated with functions from the RColorBrewer package:
total.steps$date <- gsub("^(.*)-", "", total.steps$date)
library("RColorBrewer")
blues <- colorRampPalette(brewer.pal(9,"Blues"))(12)
library("ggplot2")
hist1 <- ggplot(total.steps, aes(x = date, y = total.steps, facets = month, fill = day)) +
geom_bar(stat = "identity") + facet_grid(. ~ month) + theme_bw() + labs(title = "Total number of steps per day") + scale_fill_manual(values = blues[5:12])
hist1 + theme(axis.text.x = element_text(angle = 90, vjust = 0.5, size = 8),
strip.background = element_rect(fill = c("white")))
The mean and median of the data were calculated:
mean(total.steps$total.steps, na.rm = T)## [1] 10766
median(total.steps$total.steps, na.rm = T)## [1] 10765
The data was summarised per interval and then the mean number of steps per interval was calculated:
avg.steps <- ddply(data, .(interval), summarise, avg.steps = mean(steps, na.rm = T))Some additional processing was carried out for plotting, intervals were converted into character strings that represented a 24 hour format, "00:00":
avg.steps$time <- sprintf("%04d", avg.steps$interval)
avg.steps$time <- gsub("^([0-9]{2})([0-9]{2})$", "\\1:\\2", avg.steps$time)
avg.steps$group = c("g1") # Need to add a dummy group for plotting in ggplotA line plot was created:
x.breaks <- seq(1, 288, by = 24)
plot1 <- ggplot(avg.steps, aes(x = time, y = avg.steps, group = group)) + geom_line() +
theme_bw() + labs(title = "Average number of steps per interval")
plot1 + scale_x_discrete(breaks = avg.steps$time[x.breaks],
labels = avg.steps$time[x.breaks])
The interval with the maximum number of steps was found:
avg.steps$time[which.max(avg.steps$avg.steps)]## [1] "08:35"
The interval from 08:35 to 08:40.
Count the number of "NA" in the data:
sum(is.na(data$steps))## [1] 2304
Impute strategy, replace the NA's with the mean value for that particular interval:
impute.values <- ddply(data, .(interval), summarise, mean = mean(steps, na.rm = T))
impute <- function(x, y){
if(is.na(x)){
x <- impute.values$mean[which(impute.values$interval == y)]
}
return(x)
}
data1 <- data
data1$steps <- mapply(impute, data$steps, data$interval)The data was summarised and the total number of steps calculated. Two addtional factors were added to the dataframe, one for months and another for days of the week:
total.steps1 <- ddply(data1, .(date), summarise, total.steps = sum(steps))
total.steps1$month <- months(as.POSIXlt(total.steps1$date))
total.steps1$month <- factor(total.steps1$month, levels = c("October", "November"))
total.steps1$day <- weekdays(as.POSIXlt(total.steps1$date))
total.steps1$day <- factor(total.steps1$day, levels = c("Monday", "Tuesday", "Wednesday",
"Thursday", "Friday", "Saturday",
"Sunday"))A barplot of the data was made:
total.steps1$date <- gsub("^(.*)-", "", total.steps1$date)
hist2 <- ggplot(total.steps1, aes(x = date, y = total.steps, facets = month, fill = day)) +
geom_bar(stat = "identity") + facet_grid(. ~ month) + theme_bw() +
labs(title = "Total number of steps per day") + scale_fill_manual(values = blues[5:12])
hist2 + theme(axis.text.x = element_text(angle = 90, vjust = 0.5, size = 8),
strip.background = element_rect(fill = c("white")))
Recalculate the mean and median for the new dataframe with imputed values:
mean(total.steps1$total.steps, na.rm = T)## [1] 10766
median(total.steps1$total.steps, na.rm = T)## [1] 10766
Reworked the impute function to also group by weekend/weekday classification:
data2 <- data
data2$date <- as.POSIXlt(data2$date)
data2$day <- weekdays(data2$date)
data2$w <- ifelse(data2$day %in% c("Saturday", "Sunday"), "Weekend", "Weekday")
impute.values2 <- ddply(data2, .(interval, w), summarise, mean = mean(steps, na.rm = T))
impute <- function(x, y, z){
if(is.na(x)){
x <- impute.values2$mean[which(impute.values2$interval == y &
impute.values2$w == z)]
}
return(x)
}
data2$steps <- mapply(impute, data2$steps, data2$interval, data2$w)Ddply() was used to find the mean number of steps per interval and further grouped by weekend/weekday:
avg.steps2 <- ddply(data2, .(w, interval), summarise, avg.steps = mean(steps, na.rm = T))Made legible labels:
avg.steps2$time <- sprintf("%04d", avg.steps2$interval)
avg.steps2$time <- gsub("^([0-9]{2})([0-9]{2})$", "\\1:\\2", avg.steps2$time)Repeated the code to make a line plot of the data split by weekday/weekend:
x.breaks <- seq(1, 288, by = 24)
plot2 <- ggplot(avg.steps2, aes(x = time, y = avg.steps, group = w, colour = w,
facets = w)) + geom_line() +
facet_wrap(~ w, ncol = 1) + theme_bw() +
labs(title = "Average number of steps per interval")
plot2 + scale_x_discrete(breaks = avg.steps2$time[x.breaks],
labels = avg.steps2$time[x.breaks]) +
theme(legend.position = "none", strip.background = element_rect(fill = c("white")))
Noticeable peak observed only in the weekday data, this peak occurs between 08:00 and 09:00, suggesting travelling to work.