We’re nine days away from Makeover Monday Live at TC18! We’re expecting a full crowd of 1000 to visualize what I promise will be a fun data set. It’s also a celebration for Eva and me as we launch the Makeover Monday book with a book signing immediately after the session. We’re very excited to see all of you and celebrate what we all created as a community. If you see the session full on the app, don’t let that turn you away. Sign up for the waiting list; there are sure to be people that don’t show up.

This week’s data set was about cancer survival rates in America. I like how we’re seeing so many people progressing each week and the overall quality the past few weeks has been fantastic. People appear to be reading the lessons we write and focusing on improving their work. And that’s the whole point of Makeover Monday: Improving How We Visualize and Analyze Data, One Chart at a Time.



During viz review this week, Eva and I reviewed this beautiful piece by Michael Mixon:

One of the recommendations we made was to change the shading so that it went horizontally instead of vertically. I haven’t seen any iterations from Michael on this yet, so let me show you what I mean.

The idea we were trying to express was that the eye will naturally follow patterns that stand out. When I look at his viz, my eye goes down the page instead of across due to the shading by cancer type. However, the data goes across the page, so I was curious to see how it would look if I shaded the rows across.

Does your eye go across the page now? Shading in this manner makes it easier to see which bars go with which dimensions. Next time you want to help guide your readers around your visualization, consider using shading. Visually, this is similar to using highlighting to focus on a specific subset of the data.



Consider this chart by LM-7:

In particular, let’s have a closer look at the color legend (sorry for the blurry screenshot). When I saw this, I knew something was wrong:

These numbers simply don’t make sense to me. We’re already using percentages that can never go above 100%, so how the heck is there a percent change of 939%? It’s because LM-7 calculated the percent change from the percentage of each year compared to the 1977 percentage.

Mathematically there is nothing wrong with calculating a percent change based on percentages. The math for percent change is:

% change = (Year B – Year A)/(Year A) * 100, where  Year A and Year B are the percentages in years A and B

Using the numbers in the data set for breast cancer for all races for women, the survival rate in 1977 was 74.8% and the survival rate in 2013 was 91.1%. Calculating % change would be:

% change = (.911 – .748) /.748 * 100 = .218 * 100 = 21.8%

Again, mathematically, there is nothing wrong with this. However, I think you should consider the purpose of a percentage. Percentages have already normalized the data, so I would recommend that you look at change instead of percent change. Let’s compare calculating the percent difference from 1977 to the difference from 1977:

Using a % difference over-accentuates the change, which is misleading. LM-7 changed the viz once I pointed this out, so kudos for iterating. The key lesson is that when your base data is already percentages, use a difference instead of a percent difference to calculate change.

Now here are this week’s favorites.



AuthorAdi McCrea
Link: Twitter