Thursday, April 01, 2010

The Answer to Yesterday's Quiz!

Not really a quiz but a p.s. question in my post about the stereotype threat. I asked if anyone found something a bit off in the Wikipedia discussion of the issue.

This is what I had in mind: The piece begins

One published meta-analysis conducted by Walton & Spencer (2009) found significant evidence that stereotype threat impairs the standardized test performances of African Americans and women on the SAT.[2] However, an unpublished meta-analysis of 55 published and unpublished studies shows mixed evidence of this effect.[3]

We have one published study pointing one way and one unpublished study pointing the other way and these are given equal weight. What's wrong with that?


A recent metaanalysis by Wicherts et al.[3] states that the literature on stereotype threat is an extreme example of publication bias, or the file drawer effect. Many researchers have attempted to replicate the effect, but those studies that find positive results are more likely to be published. For every published study that finds an effect, there is another study that finds no evidence of stereotype threat, or even a negative effect (black performance higher under stereotype threat). Moreover, the few null effect studies that have been published, such as the large-scale studies by Educational Testing Service,[12] have the most statistical power and more closely resemble real testing situations. This meta-analysis, however, was rejected for publication because of numerous flaws...

My bolds.

The Wikipedia anticipates a finding which has not yet been observed! And this anticipation is given the same weight as already published work. This smells off.

It would have been perfectly fine to refer to an unpublished study, assuming a manuscript is available for those who want to read it. But to give it so much weight in the discussion, especially given that it was rejected for publication suggests to me that the writer of this article has a prior bias.'