# Using ggplot2 for Descriptive Statistics

For this we will use a new dataset:

```{r}

input <- read.csv("data/ambulance_assault.csv")

```

This contains the number of ambulance callouts to assault incidents (downloadable from the London DataStore) between 2009 and 2011.

Take a look at the contents of the file:

```{r}

head(input)

```

We can now plot a histogram to show the distribution of values.

```{r}

p <- ggplot(input, aes(x = assault_09_11))

```

Remember the `ggplot(input, aes(x=assault_09_11))` section means create a generic plot object (called p.ass) from the input object using the `assault_09_11` column as the data for the x axis. To create the histogram you need to tell R that this is what you want to go with

```{r, eval=FALSE}

p + geom_histogram()

```

The resulting message (`stat_bin: binwidth defaulted to range/30...`)

relates to the bins - the breaks between histogram blocks.

If you want the bins (and therefore the bars) to be thinner

(i.e. representing fewer values) you need to make the bins

smaller by adjusting the binwidth. Try:

```{r, eval=FALSE}

p +

geom_histogram(binwidth = 10) +

geom_density(fill = NA, colour = "black")

```

It is also possible to overlay a density distribution over the top of the histogram.

For this we need to produce a second plot object with the density distribution as the y variable.

```{r Histogram with density overlay}

p2 <- ggplot(input, aes(x = assault_09_11, y = ..density..))

p2 +

geom_histogram() +

geom_density(fill = NA, colour = "red")

```

What kind of distribution is this plot showing? You can see that there are

a few wards with very high assault incidences (over 750).

To find out which ones these are we can select them.

```{r}

input[which(input$assault_09_11>750),]

```

It is perhaps unsurprising that St James's and the West End have the highest counts.

The plot has provided a good impression of the overall distribution,

but what are the characteristics of each distribution within the Boroughs?

Another type of plot that shows the core characteristics of the distribution

is a box and whisker plot. These too can be easily produced in R

(you can't do them in Excel!). We can create a third plot object

(note that the assault field is now y and not x):

```{r}

p3 <- ggplot(input, aes(x = Bor_Code, y = assault_09_11))

```

and convert it to a boxplot.

```{r, eval=FALSE}

p3 + geom_boxplot()

```

Perhaps this would look a little better flipped round.

```{r Bar and whisker plot}

p3 + geom_boxplot() +

coord_flip()

```

Now each of the borough codes can be easily seen.

No surprise that the Borough of Westminster (00BK)

has the two largest outliers. In one line of code you

have produced an incredibly complex plot rich in information.

This demonstrates why R is such a useful program for these kinds of statistics.

If you want an insight into some of the visualisations you can develop with this type of data we can do faceting based on the example of the histogram plot above.

```{r Faceted histogram, message=FALSE}

p + geom_histogram() +

facet_wrap( ~ Bor_Code)

```

We need to do a little bit of tweaking to make this plot publishable but we want to demonstrate that it is really easy to produce 30+ plots on a single page! Faceting is an extremely powerful way of visualizing multidimensional datasets and is especially good for showing change over time.