Thursday, May 21, 2009

and now, when science goes bad

Here's an old post from the HuffPo on bad science, or what Geoffrey Dunn calls voo-doo social psychology, regarding a study that "found" that Sarah Palin's looks tanked her chances to be taken seriously. She was, according to the study, simply too sexy to be VP.

Once again (see below, on numbers) the methodology was the villain. From the post:

Let's look at the methodology itself. Heflick and Goldenberg assigned students to jot down a few lines about one of two American women celebrities, Palin or the actress Angelina Jolie. Half of the participants in each category were asked to write "your thoughts and feelings about this person," while the other half were asked to write "your thoughts and feelings about this person's appearance."

The participants were then asked to evaluate their subject (Palin or Jolie) in terms of various attributes, including "competence." Finally, they were asked to identify who they were intending to vote for in the upcoming election.

First of all, the study's sample of 133 undergraduates was hardly large enough for an accurate conclusion to be drawn. Moreover, the sample group, presumably in their late teens and early twenties, is hardly reflective of the American electorate. Generational distinctions cannot even be identified or assessed with this study. Moreover, the sample was heavily skewed toward women (96 females compared to 37 males). While party affiliations were identified, there were no variant markers for race, class or geographical origin of each participant. In purely statistical terms, its reliability is extremely low.

There's more. Read it all. bk


tk said...

Sarah Palin is not as hot as Michelle Obama

Jack said...

Did the study use a regression or ANOVA? I want to see the R-squared values!