You might have noticed that we haven't really worked with the sleep variable. Health tracker apps often ask you to round your estimated amount of sleep (even if they don't, we often round anyway). This has an interesting effect on our layered graphs.
Try layering a scatterplot of sleep over a boxplot, like we did before.
Weird—this data set has almost 30 days' worth of data, but we only see five points on our scatterplot. Are we missing some data?
Actually, because we rounded to the nearest quarter of an hour, there are only a few possible values for sleep. If we slept 8 hours on 10 different days, those data points would overlap on our graph, so that we can't see all the points at once.
In order to see all the points at once, coders can "jitter", or randomly move, the points on the graph. If a scatterplot was created by someone carefully placing the points where they are supposed to be, a jitterplot is what happens when we come along and shake things up.
Let's recreate our layered boxplot and scatterplot of sleep, using gf_jitter() instead of gf_point().
Another option we could use to reveal points that are hidden behind each other is to use special arguments.
Remember how we used the bins argument to adjust our histograms? bins is only for histograms, but all graphs can be adjusted using additional arguments called aesthetics. For example, alpha controls transparency using values from 0 to 1. size controls the size of the points.
Try using a combination of these to improve our scatterplot in our first scatterplot + boxplot graph. Why is this a better representation of our data?
<iframe data-type="datacamp" id="extra-8-3" height="350" src="https://uclatall.github.io/mtucker-coding-study/data-camp/dc-extra-8-3.html"></iframe>