How to use color to see your numbers

The best way to present data is where the conclusion is clear. If you need to still interpret the data for your audience, you may not have done a good enough job. One of the ways to do this is coloring values based on a scale. I particularly like the jet colormap scheme, but more importantly, the tools to color data and improve visualization are widely available and easy to use.

I have been using Microsoft Excel for close to 20 years, and even though I have mostly switched to Numbers, the one key functionality that I still love about Excel is Conditional Formatting. This allows a coloring to be used for a large set of data, and it helps to visually see trends rather than digging into each value. Typically, this is done by plotting data, but some times, even a regular plot doesn't quite capture all the nuances of the data. Conditional formatting allows this to be done over a large amount of data, quickly, and creatively.

First thing, get some data. I picked some surface temperature data (https://data.giss.nasa.gov/gistemp/). It's an interesting dataset because it is the difference in temperature relative to the average of the temperatures from 1951 to 1980. The reported values are in celsius, so multiply by 1.8 to get the changes in Fahrenheit. The data is per month per year. Some people have interpreted the data one way or another, and my plan is to only present the data.

Firstly, let's look at just a snippet:

That is a lot of numbers. Ugh. Let's add some color based on all the values:

That's a little better, but it is hard to understand the trend because each month has a different trend. Let's color based on each month, so each month will define the minimum and maximum for that color map:

Much better. What about a different color scheme using white instead of yellow as a middle color:

Let's modify the color to have a minimum threshold and maximum threshold:

Now, let's look at some plots. The first is time linear, but there is a lot of noise because of the number of samples. The second is a box plot per year:

Now, let's get weird and add some crazy color:

Let's make it a surface:

Both of these could still be difficult to understand right away. So let's plot all the data in something more manageable:

Let's color based on column:

What happens if we make a video? Would it provide more impact to the conclusion of the data?

Usually, I end up with multiple versions of the same data, presented in different ways before I present. I then try to put myself in the shoes of another, and I remember that I am burdened with knowledge, so the way I look at my graphs is different than the way another would. That's why, I'll end with a plot where I only changed the temperature differences from Celsius to Fahrenheit: