Dataviz

Adding Standard Deviation to a Dataviz

Adding Standard Deviation to a Dataviz

Do you need to communicate your standard deviations to your audience? No, I mean, really DO YOU? Think hard about that because chances are your audience doesn’t give a crap. They want to know you calculated your standard deviations. They want to know you were meticulous in your number crunching. But most likely they do not want you to actually tell them the standard deviation. Tuck that ish in an appendix or something.

But in the rare circumstance where your audience really does want to see your standard deviations, here’s a pretty easy way to visualize them in your graph inside Excel. It represents the standard deviation via a gray band behind the mean scores, using a stacked area chart. Sounds complicated, I know, but it’s actually pretty simple and these are the step-by-step instructions. In this example, we are visualizing mean reading scores for schools across the state.

Step 1, we calculate means and standard deviations for each of the 5 years we are reporting on.

StdDev

 

Step 2, add the standard deviation to the mean to calculate the upper end of the standard deviation range. And subtract the standard deviation from each mean to get the lower end of the standard deviation range.StdDev2

Step 3, the difference between the upper range and the lower range is the full standard deviation range, so calculate that by subtracting the lower range from the upper range.StdDev3

Step 4, graph. We want to create a stacked area chart, using this guy:

StdDevType

So highlight the values in the lower range and the full range, click the stacked area graph, and get this:

StdDevStacked

 

As my notes suggest, change the default colors of the stacked areas so that the lower range is no fill or white and the actual standard deviation range is gray. That’s your gray background, yo!

Step 5, add the mean values. First, right-click on the graph, click on Select Data, and add your mean values as a new series. It’ll look like this:

StdDev6

You don’t want it to look like that, but don’t freak out. Just keep going.

Step 6, click on that new area representing your means so that it is highlighted. Then change the chart type and select a line.

StdDev8

It’ll look like this:

StdDev9

Whew! That’s more like it.

Step 8, format! Add an explanatory title and a legend-like subtitle to convey what standard deviation actually means.StdDev10And there you have it. But wield this with care, okay? Standard deviation, while important to us datanerds, can elicit deer-in-the-headlight stares from many audiences.

7 thoughts on “Adding Standard Deviation to a Dataviz
  1. Nic Bookman says:

    Subscribing to the blog!

  2. Kristen says:

    Thanks for sharing, I am subscribing now!

  3. Jen Peterson says:

    Fantastic – thanks so much! Perfect timing too – we are just producing our final evaluation report.

  4. kirsty wright says:

    Elegant and clear

  5. David N says:

    Here is a link to a trend chart that shows CIs (confidence intervals) for each segment of the data. Very similar in principle to showing SD (obviously).

    https://twitter.com/Biff_Bruise/status/446633864908136448

    I did this one back in 2011… look familiar? :)

  6. Angie Ficek says:

    this is great, stephanie! i hadn’t thought of using an area graph to do this, especially for data over time. i’ve been messing around with line graphs / adding hi-low lines to show standard deviations. will try the area graph next time!

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

@KZOODISTILLING the people of kalamazoo demand details.

RT @RadPresenters: We are recording our next episode soon & we'll block out time to answer your Qs. What are you struggling with? Send ques…

Looking for an embeddable survey tool that will let respondents use a slider to position themselves on a spectrum. Know of any?

Check out the expanded #dataviz section of @BetterEval that we've been cooking up! http://t.co/Qf85ZhiNgk #eval

Afternoon friends: The rules about when it's ok to NOT start your axis at zero. http://t.co/jDOKX3lP2o #dataviz

RT @esjewett: @evergreendata Yes! Also when the traditional 0-value is no more meaningful than any other value. Temperature, for instance.

RT @AnnKEmery: [New post] Span Charts: When you've only got the min and max, not a mean/median/frequency http://t.co/WQeKFvp0RF http://t.co…

Perhaps a controversial post: When It's Ok to NOT Start your Axis at Zero http://t.co/SdDeI7sHqP #dataviz

RT @parisakharazi: Useful find: Introducing the Data Visualization Checklist - http://t.co/oYPunrz33g Thanks for the great tips! @evergreen