Activity Monitoring Analysis
Loading and preprocessing the data
- Firstly, let’s read csv file from the web:
#Loading library
library(plyr)
library(ggplot2)
library(gridExtra)#Creating temporary file in memory
temp <- tempfile()
#Just to use download.file in R Markdown
setInternet2(use = TRUE)
#Download zip file in temp
download.file("https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2Factivity.zip", temp)
#Unzip temp file and read "activity.csv" from it to activity variable
activity <- read.csv(unz(temp, "activity.csv"))
#Delete temp file
unlink(temp)- Explore size of data.frame:
dim(activity)[1] 17568 3- Explore type of variables in activity data.frame:
str(activity)'data.frame': 17568 obs. of 3 variables:
$ steps : int NA NA NA NA NA NA NA NA NA NA ...
$ date : Factor w/ 61 levels "2012-10-01","2012-10-02",..: 1 1 1 1 1 1 1 1 1 1 ...
$ interval: int 0 5 10 15 20 25 30 35 40 45 ...So, steps variable is correct type. It is better to change type of date to some date type. interval variable represents minutes interval. It is better to transform this column to two columns, one of which will represent start hour, second - start minute.
#Tranforming date variable to Date type
activity$date <- as.Date(activity$date)
#Adding two variables based on interval variable
activity$interval <- as.character(activity$interval)
activity$interval <- sapply(activity$interval, FUN = function(element) return(paste(as.character(c(rep(0, (4-nchar(element))), element)), collapse = "")))
activity$start_hour <- as.integer(sapply(activity$interval, FUN = function (element) return (substr(element, 1, 2))))
activity$start_minute <- as.integer(sapply(activity$interval, FUN = function (element) return (substr(element, 3, 4))))Now, data looks like:
head(activity) steps date interval start_hour start_minute
1 NA 2012-10-01 0000 0 0
2 NA 2012-10-01 0005 0 5
3 NA 2012-10-01 0010 0 10
4 NA 2012-10-01 0015 0 15
5 NA 2012-10-01 0020 0 20
6 NA 2012-10-01 0025 0 25And type of variables:
str(activity)'data.frame': 17568 obs. of 5 variables:
$ steps : int NA NA NA NA NA NA NA NA NA NA ...
$ date : Date, format: "2012-10-01" "2012-10-01" ...
$ interval : chr "0000" "0005" "0010" "0015" ...
$ start_hour : int 0 0 0 0 0 0 0 0 0 0 ...
$ start_minute: int 0 5 10 15 20 25 30 35 40 45 ...What is mean total number of steps taken per day?
- Total number of steps per day:
total_steps <- ddply(activity, ~date, summarise, Total_steps=sum(steps, na.rm = TRUE))
total_steps date Total_steps
1 2012-10-01 0
2 2012-10-02 126
3 2012-10-03 11352
4 2012-10-04 12116
5 2012-10-05 13294
6 2012-10-06 15420
7 2012-10-07 11015
8 2012-10-08 0
9 2012-10-09 12811
10 2012-10-10 9900
11 2012-10-11 10304
12 2012-10-12 17382
13 2012-10-13 12426
14 2012-10-14 15098
15 2012-10-15 10139
16 2012-10-16 15084
17 2012-10-17 13452
18 2012-10-18 10056
19 2012-10-19 11829
20 2012-10-20 10395
21 2012-10-21 8821
22 2012-10-22 13460
23 2012-10-23 8918
24 2012-10-24 8355
25 2012-10-25 2492
26 2012-10-26 6778
27 2012-10-27 10119
28 2012-10-28 11458
29 2012-10-29 5018
30 2012-10-30 9819
31 2012-10-31 15414
32 2012-11-01 0
33 2012-11-02 10600
34 2012-11-03 10571
35 2012-11-04 0
36 2012-11-05 10439
37 2012-11-06 8334
38 2012-11-07 12883
39 2012-11-08 3219
40 2012-11-09 0
41 2012-11-10 0
42 2012-11-11 12608
43 2012-11-12 10765
44 2012-11-13 7336
45 2012-11-14 0
46 2012-11-15 41
47 2012-11-16 5441
48 2012-11-17 14339
49 2012-11-18 15110
50 2012-11-19 8841
51 2012-11-20 4472
52 2012-11-21 12787
53 2012-11-22 20427
54 2012-11-23 21194
55 2012-11-24 14478
56 2012-11-25 11834
57 2012-11-26 11162
58 2012-11-27 13646
59 2012-11-28 10183
60 2012-11-29 7047
61 2012-11-30 0- Histogram of total number of steps:
rng <- range(total_steps$Total_steps)
p1 <- qplot(Total_steps,
data=total_steps,
geom="histogram",
breaks = seq(rng[1], rng[2], by =1700),
main = "Histogram of number of\n steps per day",
xlab = "Steps per day",
ylab = "Days"
)
p1
- Mean and median of the total number of steps taken per day:
#Mean number of total number of steps per day
mean(total_steps$Total_steps, na.rm = TRUE)[1] 9354.23#Median number of total number of steps per day
median(total_steps$Total_steps, na.rm = TRUE)[1] 10395What is the average daily activity pattern?
- Average number of steps per 5 minute interval:
activity$minute_from_00 <- activity$start_hour * 60 + activity$start_minute
avr_steps_5min <- ddply(activity, ~minute_from_00, summarise, Avr_steps=mean(steps, na.rm = TRUE))
head(avr_steps_5min) minute_from_00 Avr_steps
1 0 1.7169811
2 5 0.3396226
3 10 0.1320755
4 15 0.1509434
5 20 0.0754717
6 25 2.0943396- Plotting average values per 5 minutes:
p2 <- qplot(minute_from_00,
Avr_steps,
data=avr_steps_5min,
geom=c("point", "path"),
main = "Average of number of\n steps per 5 minute interval",
xlab = "Minutes from 00:00",
ylab = "Average number of steps"
)
p2
- 5 minutes interval with max number of average steps:
avr_steps_5min[which.max(avr_steps_5min$Avr_steps),] minute_from_00 Avr_steps
104 515 206.1698So, max number of steps at average is taken place from 08:35 to 08:40.
Imputing missing values
- Total number of missing values in the dataset:
#Number of NAs in activity data.frame
apply(activity[,1:3], 2, FUN = function(column) return(sum(is.na(column)))) steps date interval
2304 0 0 So, there are 2304 NAs in steps column and no NAs in date and interval columns.
- Replacing NAs in steps column with average number of steps per 5 minute interval:
activity_noNA <- activity
for (i in 1:nrow(activity_noNA)) {
if (is.na(activity_noNA$steps[i])) {
activity_noNA$steps[i] <- avr_steps_5min$Avr_steps[avr_steps_5min$minute_from_00 == activity_noNA$minute_from_00[i]]
}
}
#Number of NAs in activity_noNA data.frame
apply(activity_noNA[,1:3], 2, FUN = function(column) return(sum(is.na(column)))) steps date interval
0 0 0 There are no NAs left.
Head of the data now looks like:
head(activity_noNA) steps date interval start_hour start_minute minute_from_00
1 1.7169811 2012-10-01 0000 0 0 0
2 0.3396226 2012-10-01 0005 0 5 5
3 0.1320755 2012-10-01 0010 0 10 10
4 0.1509434 2012-10-01 0015 0 15 15
5 0.0754717 2012-10-01 0020 0 20 20
6 2.0943396 2012-10-01 0025 0 25 25- Total number of steps taken each day, calculated based on data with imputed NA:
total_steps_noNA <- ddply(activity_noNA, ~date, summarise, Total_steps=sum(steps, na.rm = TRUE))
head(total_steps_noNA) date Total_steps
1 2012-10-01 10766.19
2 2012-10-02 126.00
3 2012-10-03 11352.00
4 2012-10-04 12116.00
5 2012-10-05 13294.00
6 2012-10-06 15420.00Two histogram of total number of steps without and with imputed NAs:
rng <- range(total_steps_noNA$Total_steps)
p3 <- qplot(Total_steps,
data=total_steps_noNA,
geom="histogram",
breaks = seq(rng[1], rng[2], by =1700),
main = "Histogram of number of\n steps per day with imputed NA",
xlab = "Steps per day",
ylab = "Days")
grid.arrange(p1, p3, ncol=2)
- Mean and median of the total number of steps taken per day without and with imputed NAs:
#Mean number of total number of steps per day
mean(total_steps$Total_steps, na.rm = TRUE)[1] 9354.23mean(total_steps_noNA$Total_steps)[1] 10766.19#Median number of total number of steps per day
median(total_steps$Total_steps, na.rm = TRUE)[1] 10395median(total_steps_noNA$Total_steps)[1] 10766.19Imputing has next impact:
1. Histogram changed its view a little bit. Now, there are no days with no steps, but more days with ~11000 steps. It as because we changed total number of steps for 8 days;
2. Mean and Medium values of steps became the same -> distribution became more normal.
Are there differences in activity patterns between weekdays and weekends?
- Creating a new factor variable with weekday and more for identifying weekend:
Sys.setlocale("LC_TIME", "English")[1] "English_United States.1252"activity_noNA$weekday <- factor(weekdays(activity_noNA$date))
activity_noNA$weekend <- factor(ifelse((activity_noNA$weekday == "Saturday" | activity_noNA$weekday == "Sunday"), "weekend", "weekday"))- Creating data for weekdays and weekends:
avr_steps_5min_weekends <- ddply(activity_noNA, .(minute_from_00, weekend), summarise, Avr_steps=mean(steps))
head(avr_steps_5min_weekends) minute_from_00 weekend Avr_steps
1 0 weekday 2.25115304
2 0 weekend 0.21462264
3 5 weekday 0.44528302
4 5 weekend 0.04245283
5 10 weekday 0.17316562
6 10 weekend 0.01650943Visualizing with one plot, but two series (it is the same as using panel plots):
qplot(minute_from_00,
Avr_steps,
color = weekend,
data=avr_steps_5min_weekends,
geom=c("point", "path"),
main = "Average of number of\n steps per 5 minute interval",
xlab = "Minutes from 00:00",
ylab = "Average number of steps"
)
Very interesting pattern could be observed: during weekdays average number of steps is higher for morning hours, where for weekends - day hours. For the evening hours - patterns are very similar, but ~ from 19:00 till 21:00 average number of steps is higher for weekends.