We created vertical, horizontal, and stacked bar charts using the barplot(...) function and its custom arguments. We also expanded upon the legend(...) function to gain more control over its appearance. Let us reflect upon each of these steps.

We started by adding text labels to our bars via the names argument. This argument receives a vector containing the text label to be appended to each bar. In our case, the labels consisted of the four battle methods that follow:

barAllMethodsSoldiersNames <- c("Fire", "Ambush", "Head to Head",
"Surround")

Then, we looked at two arguments that are unique to the barplot(...) function. The width argument specifies the thickness of a chart's bars. It can be defined as a single value for all bars or a vector that contains unique values for each bar. Note that if a single value is used, the xlim argument must be defined for the width argument to take effect. In coordination with width, we also employed the space argument, which determines the distance between a chart's bars. Like width, space can be defined as a single value or a vector containing values for each bar. It is measured as a ratio of the average bar width and defaults to a value of 0.2. For example, if the average width of the bars was 5 and the space was set to 0.5, then the distance between each bar would be 2.5. For our chart, we chose a width of 0.25 and a space of 2, which had the visual effect of making our bars skinnier and spread farther apart:

> barAllMethodsDurationWidth <- 0.25
> barAllMethodsDurationSpace <- 2

We chose uniform width and space values for our bars. Had we wanted to set unique values for each bar, such as when weighting the bars according to the number of data points they include, we could have used the following code:

> barAllMethodsDurationWidth <- c(0.1, 0.25, 0.5, 0.75)
> barAllMethodsDurationSpace <- c(0.5, 1, 1.5, 2)

Lastly, the names, width, and space arguments were incorporated into our overall barplot(...) function:

> barplot(height = barAllMethodsDurationBars,
main = barAllMethodsDurationLabelMain,
xlab = barAllMethodsDurationLabelX,
ylab = barAllMethodsDurationLabelY,
xlim = barAllMethodsDurationLimX,
ylim = barAllMethodsDurationLimY,
col = barAllMethodsDurationRainbowColors,
names = barAllMethodsDurationNames,
width = barAllMethodsDurationWidth,
space = barAllMethodsDurationSpace)

To further expand our bar chart, we incorporated the horiz argument. This argument allows us to reorient our bars such that they extend horizontally, rather than vertically, across the chart. The horiz argument receives either a TRUE or FALSE value indicating whether the bars should be oriented horizontally. By default, horiz is set to FALSE and the bars are drawn vertically. We chose to reorient our bars by setting our horiz variable to TRUE:

barAllMethodsDurationHoriz <- TRUE

We then included it into our barplot(...) function:

> barplot(height = barAllMethodsDurationBars,
main = barAllMethodsDurationLabelMain,
xlab = barAllMethodsDurationLabelY,
ylab = barAllMethodsDurationLabelX,
xlim = barAllMethodsDurationLimY,
ylim = barAllMethodsDurationLimX,
col = barAllMethodsDurationRainbowColors,
names = barAllMethodsDurationNames,
width = barAllMethodsDurationWidth,
space = barAllMethodsDurationSpace,
horiz = barAllMethodsDurationHoriz)

Note that reorienting the bars of a chart is similar in effect to rotating it by 90 degrees. Therefore, to prevent a misshapen and unreadable graphic, we must also swap all arguments related to the x and y-axes. In our case, that meant exchanging our x-axis and y-axis limits and labels:

> barplot(height = barAllMethodsDurationBars,
main = barAllMethodsDurationLabelMain,
xlab = barAllMethodsDurationLabelY,
ylab = barAllMethodsDurationLabelX,
xlim = barAllMethodsDurationLimY,
ylim = barAllMethodsDurationLimX,
col = barAllMethodsDurationRainbowColors,
names = barAllMethodsDurationNames,
width = barAllMethodsDurationWidth,
space = barAllMethodsDurationSpace,
horiz = barAllMethodsDurationHoriz)

After swapping these values, the chart displays appropriately. Had we forgotten to make this exchange, we would have ended up with the following graphic:

We then turned to developing a new chart that would make use of the beside argument. This argument tells a chart's bars to stack atop one another, rather than stand side by side. Like horiz, beside accepts a TRUE or FALSE value. If TRUE, the default setting, the bars will display side by side. If FALSE, the bars will be stacked. We chose to stack our bars by setting beside to FALSE.

barAllMethodsSoldiersBeside <- FALSE

Note that for the beside argument to take effect, the height argument must be in matrix form. To organize our data into a matrix, we used the matrix(...) function, whose basic format is as follows:

matrix(values, rows, columns)

Here, values is a vector containing the relevant data points, rows is the number of rows, and columns is the number of columns. Our matrix consisted of eight data points that were organized into two rows (Shu and Wei) and four columns (Fire, Ambush, Head to Head, and Surround). In the final chart, the four bars are formed by stacking the two rows within each column; the following is the code:

> barAllMethodsSoldiersBars <- matrix(c(meanShuSoldiersFire,
meanWeiSoldiersFire, meanShuSoldiersAmbush, meanWeiSoldiersAmbush,
meanShuSoldiersHeadToHead, meanWeiSoldiersHeadToHead,
meanShuSoldiersSurround, meanWeiSoldiersSurround), 2, 4)

Our stacked bar chart depicted the average number of soldiers that are engaged in each type of battle. By stacking our bars, we were able to specify what proportion of the soldiers came from the Shu and Wei armies. Thus, our chart was able to include more information in the same amount of space:

> barplot(height = barAllMethodsSoldiersBars,
main = barAllMethodsSoldiersMain,
xlab = barAllMethodsSoldiersLabX,
ylab = barAllMethodsSoldiersLabY,
names = barAllMethodsSoldiersNames,
beside = barAllMethodsSoldiersBeside)

After beside, we used the density and angle arguments to define the shading of our bars. The density argument defines the closeness of the shaded lines. It receives either a single non-negative value for all matrix rows or a vector that contains values for each row. The angle argument specifies the angle at which the shaded lines are to be drawn. It also accepts a single value for all matrix rows or a vector containing values for each row.

Our stacked bar chart used a density of 10 for the Shu row and 25 for the Wei row:

> barAllMethodsSoldiersDensity <- c(10, 25)

It also featured an angle of 45 for the Shu and -45 for the Wei:

> barAllMethodsSoldiersAngle <- c(45, -45)

Hence, you will notice that the shading in the Shu portions of our bars is spread thinner and rises to the upper-right of the chart, whereas the shading in the Wei portions of the chart is thicker and declines towards the lower-right of the chart:

> barplot(height = barAllMethodsSoldiersBars,
main = barAllMethodsSoldiersMain,
xlab = barAllMethodsSoldiersLabX,
ylab = barAllMethodsSoldiersLabY,
names = barAllMethodsSoldiersNames,
beside = barAllMethodsSoldiersBeside,
density = barAllMethodsSoldiersDensity,
angle = barAllMethodsSoldiersAngle)

In the final step, we added a legend to our chart. A legend is critical to a stacked bar chart, because it indicates the difference between its grouped regions. Our legend(...) function expanded upon the legends that we created in the previous chapter. We positioned the legend towards the upper-left side of the chart using the x and y arguments. We also specified the labels that we wanted to show in the legend (Shu and Wei). By default, the legend would have displayed the bar names (Fire, Ambush, Head to Head, and Surround). Since we needed to display the stacked segments of each bar instead, we had to specifically define them as a vector in the legend argument. Next, we incorporated the exact density and angle arguments from our barplot(...) function. This matched the legend's shading to that of our chart. To complete our legend, we used the cex argument to multiply its size by 2 times. The cex argument accepts a numeric value that indicates how much a legend should be scaled by. Increasing the size of our legend made it easier to read, thus enabling viewers to quickly distinguish between our chart's stacked regions:

> legend(x = 0.2, y = 70000, legend = c("Shu", "Wei"),
density = barAllMethodsSoldiersDensity,
angle = barAllMethodsSoldiersAngle, cex = 2)

Use your soldiersByCity dataset to create a chart that depicts the total number of soldiers in the Shu and Wei armies as two separate bars. Then create a stacked bar chart with the same data, but separate the Shu and Wei bars into distinct sections for each city. Compare these two charts and reflect upon the pros and cons of using each.