Different Scales

comparing distributions with multiple Density Plots

We discussed density plots for describing the distribution of a single variable on sec-uni-density. Overlapping multiple density plots is an effective way of comparing multiple distributions. We will treat this as a solution to the deficiencies of plotting multiple histograms on page page sec-multiple-histograms.

Time series with very different scales

This is another great example of how to deal with a difficult problem in scales. Here, the issue is that…

PH Fig 1

You may be tempted to make a double y axis – don’t do it!

PH Fig 2

What are some alternatives? We can normalise the values to a standard range (min = 0, max = 1):

PH Fig 3

now we can try to facet again, it looks more interesting this time:

PH Fig 4

But of course, the advantage of having the two series on the same scale means that we can have them both on a single y axis:

Colour according to trend, this one I like, very clear trend here, not as fancy as the colours, above, but intuitive. You can also add grey rectangles in the backtround marking the days.

PH Fig 5

We can also break apart the larger series, and plot each day separately.

PH Fig 6

It’s pretty noise data, so we may want to add a smooth:

PH Fig 7