PA1_template.Rmd

# Activity Statistics

## Getting the data

**<span style="color:blue">1 - Code for reading in the dataset and processing the data</span>**

Let's first download and read the data into a data frame

```{r read}
temp <- tempfile()
download.file("https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2Factivity.zip",temp)
activity <- read.csv(unz(temp, "activity.csv"))
unlink(temp)
```

Some fields have undesired data format:

```{r str}
str(activity)
```

Let's convert *interval* variable into factor and create a variable *hour* based on intervals:

```{r reformat}
library(plyr)
activity$hour<-as.factor(round_any((as.numeric(activity$interval)+1)/100, 1, f = ceiling)) # create a variable for each hour
activity$interval<-as.factor(activity$interval)
```

## Exploratory Data Analysis

**<span style="color:blue">2 - Histogram of the total number of steps taken each day</span>**
```{r stepseachday}
library(ggplot2)
ggplot(data = activity,
       aes(as.Date(as.character(activity$date), "%Y-%m-%d"), steps)) +
        stat_summary(fun.y = sum, # adds up all observations
                     geom = "bar") +
        xlab ("Dates") +
        ggtitle("Total number of steps taken each day")
```

**<span style="color:blue">3 - Mean and median number of steps taken each day</span>**

Let's create a data frame with means and medians for each day (keeping NAs for now)
```{r means_medians}
means<-ddply(activity, "date", summarise, mean = mean(steps))
medians<-ddply(activity, "date", summarise, median = median(steps))
means_medians<-merge(means,medians)
```

All medians are 0:

```{r print2, echo=FALSE}
print(means_medians)
```

That happens because there are no steps in more than 50% of intervals.

**<span style="color:blue">4 - Time series plot of the average number of steps taken</span>**

Let's visualize avarage number of steps taken (ignoring NAs for now):

```{r timeseries}
ggplot(means_medians, aes(y=mean, x=as.Date(as.character(means_medians$date), "%Y-%m-%d"))) + 
        geom_bar(stat='identity') +
        ylab ("Average daily steps") +
        xlab ("Dates") +
        ggtitle("Average number of steps taken")
```        
        
**5 - The 5-minute interval that, on average, contains the maximum number of steps</span>**

```{r meansbyintervals}
meansbyintervals<-ddply(activity, "interval", summarise, mean = mean(steps,na.rm = TRUE)) 
meansbyintervals<-arrange(meansbyintervals, desc(mean))
meansbyintervals[1,]
```

The most walkable interval is 8:35am

## Navigating Missing Values

**<span style="color:blue">6 - Code to describe and show a strategy for imputing missing data</span>**

Let's see at the missing values
```{r}
NAs <- is.na(activity$steps)
sum(NAs, na.rm = T)
mean(NAs, na.rm = T)
```

13% is a rather high number. Let's see if there any particular days or intervals where there are missing values:

```{r}
dates <- activity$date
dates <- as.Date(as.character(dates), "%Y-%m-%d")
hist(dates[NAs], "day", , xlab = "Date"
                , main = "Histogram: what dates are missing")
```

Sounds like that missing values are spread  between 8 out of 61 days. Not too bad. Let's remove these observations:

```{r}
completeactivity <- activity[complete.cases(activity), ]
```

**<span style="color:blue">7 - Histogram of the total number of steps taken each day after missing values are imputed</span>**

The plot obviously doesn't change much:

```{r}
ggplot(data = completeactivity,
       aes(as.Date(as.character(completeactivity$date), "%Y-%m-%d"), steps)) +
        stat_summary(fun.y = sum,
                     geom = "bar") +
        xlab ("Dates") +
        ggtitle("total number of steps taken each day after missing values are imputed")
```
                     
**<span style="color:blue">8 - Panel plot comparing the average number of steps taken per 5-minute interval across weekdays and weekends</span>**

Let's first do some formatting and define weekdays and weekends

```{r}
library(timeDate)
completeactivity$date<-as.Date(as.character(completeactivity$date), "%Y-%m-%d")
completeactivity$weekday<-isWeekday(completeactivity$date)
completeactivity$weekday[completeactivity$weekday==TRUE] <- "Weekday"
completeactivity$weekday[completeactivity$weekday==FALSE] <- "Weekend"
completeactivity$weekday<-as.factor(completeactivity$weekday)
```

Create a new data frame with means by weekdays and weekends and by hourly intervals:

```{r}
completeactivitymeans<-aggregate(completeactivity$steps
                                 , by=list(completeactivity$hour,completeactivity$weekday)
                                 , FUN=mean)
names(completeactivitymeans)<-c("hour","day","steps")
```

Finally, build a plot of means:

```{r}
ggplot(data = completeactivitymeans,
       aes(hour, steps)) + 
        geom_point(aes(color=hour)) +
        facet_grid(day~.) +
        ggtitle("Means by hourly intervals")
```

And other statistics:

```{r}
ggplot(data = completeactivity,
       aes(hour, steps)) + 
        geom_boxplot(aes(color=hour)) +
        facet_grid(weekday~.) +
        ggtitle("Statistics by hourly intervals")
```

## Summary

**Weekday and Weekend patters are different**

- Wakes up later on weekends

- Average amount of steps per interval is higher during weekends

- Walking on weekends is more popular

**Other data caviats and findings**

- 8 days are missing oobeservations (forgot to track?)

- The excercise showed that exploring daily aggregated or hourly aggregated data would make more sense - interval data is skewed because there are a lot of 0 steps as well as intervals with very high numbers of steps

**Thank you for the review!**