Monday, December 22, 2014

Justin Wolfers' "A Persuasive Chart Showing How Persuasive Charts Are"

Wolfers (writing for the New York Times) summarizes a study from Wansink and Tal (2014) in which participants were either a) presented with just in-text data about a drug trial or b) the text as well as with a bar graph that conveyed the exact same information. The results can be read below:


According to Wansink and Tal, the effects seems to be strongest in people who agreed with the statement "I believe in science". So, a graph makes a claim more "sciencier" and, therefore, more credible? Also, does this mean that science believers aren't being as critical because they already have an underlying belief in what they are reading? 

I thinks this is a good way of conveying the power of graphs to students in a statistics class as well as the need for better scientific literacy/statistical consumerism/skepticism. I think it is also a good example of Elaboration Likelihood Model

Finally, can we take just one moment to discuss the name of the original research article, Blinded with science: Trivial graphs and formula increase ad persuasiveness and belief in product efficacy. I'm always a fan of the "<funny cultural reference>: <serious science-y stuff>" naming convention oft used in scientific articles, and this is a real gem.

Monday, December 15, 2014

Kristopher Magnusson's "Interpreting Cohen's d effect size"

Kristopher Magnusson (previously featured on this blog for his interactive illustration of correlation) also has a helpful illustration of effect size. While this example probably has some information that goes beyond an introductory understanding of effect size (via Cohen's d) I think this still does a great job of illustrating how effect size measures, essentially, the magnitude of the difference between groups (not how improbably those differences are). See below for a screen shot of the tool., created by Kristopher Magnusson

Wednesday, December 10, 2014

UCLA's "What statistical analysis should I use?"

This resource from UCLA is, essentially, a decision making tree for determining what kind of statistical analysis is appropriate based upon your data (see below).
Screen shot from "What statistical analysis should I use?"
Now, such decision making trees are available in many statistics text book...however...
what makes this special is the fact that with each test comes code/syntax as well as output for SAS, Stata, SPSS, and R. Which is helpful to our students (and, let's be honest, us instructors/researchers as well).

Monday, December 1, 2014

Tessa Arias' "The Ultimate Guide to Chocolate Chip Cookies"

I think this very important cookie research is appropriate for the Christmas cookie baking season. I also believe that it provides a good example of the scientific method.

Arias started out with a baseline cookie recipe (baseline Nestle Toll House Cookie Recipe, which also served as her control group) and modified the recipe in a number of different ways (IVs) in order to study several dependent variables (texture, color, density, etc.). The picture below illustrates the various outcomes per different recipe modifications.

For science!

Also, being true scientist, her original study lead to several follow up studies investigating the effect of different kinds of pans and flours upon cookie outcomes.
I used this example to introduce hypothesis testing to my students. I had them identify the null and alternative hypotheses, the control and experimental groups, etc.