Assist Not Restrain Interpretation - Visualize Your Variation
By: Kristin Hunter-Thomson
YOU WANT ME TO ADD WHAT TO MY CHARTS?
Depending on what school you teach in or grade level you are working with, you may have come across the somewhat dreaded reality of visualizing variation in graphs.
Any time data are presented in the primary literature scientists visualize the variability in a range of ways, they: plot confidence intervals around raw data, include standard error bars on averages, etc. This is a fundamental part of communicating with data in science, as the variation of our data gives us LOTS of information about our claims from the data.
Great, so we need to visualize our variation in science. If you are like me, and maybe I am just alone here, when it comes to getting kids to visualize the variation things start to fall apart.
Either my students just use the defaults in Excel, neither of which make sense, without having any ideas of what they are actually plotting:
Or follow through the step-by-step instructions that I provide for Excel do it for them correctly (in that the standard deviation of the data is actually what is used as the length of the error bars - NOTE these data only had standard deviation calculated, this can also be done with standard error values)
OR they look at me like I have grown four heads when I even say "error bars."
LET'S STEP BACK TO THE "WHY"
Before we dive into the nitty gritty of actually how to visualize variation (aka how to get them to the middle option above in Excel), let's think about why we visualize variation in our data.
The simple answer = to understand it better!
Knowing how much our data varies helps us better understand what the pattern is in the data and how confident we are in the pattern we are seeing. Meaning, are the treatment and control groups really different or not, is there actually a pattern over time or not, etc.
RETHINKING THE "HOW"
So, let's use that as our backbone as we think about how to visualize variation. The whole point is to help with interpretation. And if that is the point, are those little vertical lines really helping our students?
I say not really.
Stephanie Evergreen in "Adding Standard Deviation to a Dataviz" talks through a great way to rethink how we visualize our error bars, as demonstrated here with the same data as used above.
What stands out about plotting these data (standard deviation around the mean) as a swath behind the data is that it makes more sense in how to interpret the data. The grey bar helps in your interpretation of the data!
You are able to more intuitively see the range of reading scores over time and how that whole range of scores has been increasing over time. Rather than individual bars positioned vertically at each yearly average (as above), this provides the information (standard deviation) in a way that includes broader context.
When we know that our students struggle out of the gates with understanding variation and what to do with it, maybe spending a bit of time thinking outside the box of how we visualize or variability could help!
Try it, see what happens, and let us know what you think.
LOOKING FOR MORE?
Looking for more about how to talk about variability in your data while teaching? Check out the "Identify & Explore Variability in Your Data" resources.
Looking for more ideas of how to think about visualizing variability differently? Check out "Charting Confidence Intervals" by Angie Ficek
