Quote:
Originally Posted by filtherton
There are many instances where biases in self reported studies would mostly go in one direction. For instance, women who have children with birth defects are more likely to report that they were exposed to chemicals during their pregnancies than women whose children were born without birth defects. There is also the classic example of polls over-estimating support for minority candidates because poll takers don't want to seem racist.
That's why it would be nice to see the actual study- you've got access dippin, don't you? A quote from one of the researchers showed that they were at least aware of the considerable pressure a woman may be under to conjure up a g-spot she may or may not actually have. It would be informative to see how the researchers dealt with this potential source of bias in their survey. It might also be informative to see how the prevalence of g-spots reported by the participants compares to g-spot prevalence reported in other studies.
Just for the record, one of the reasons I can't wait to get started in grad school is that I will get access to pubmed back.
|
But when there is systematic bias like that, that becomes an included variable, and tests on omitted variables turn significant.
I'm not saying self reported studies are perfect, just that they are informative and that there are several known ways to deal with the most obvious statistical issues.
Regarding the issue you mentioned, they include a number of variables in the study, some related to individual experience, some to the so called "measurement error" and that is how the eliminate the "genetic" aspect of it all. For some bias to explain the insignificance of genetic factors in a multivariate study like this, the bias would have to be one that is extremely high correlated with experience and uncorrelated with genetics. Is it possible to have something like that? Sure, but that is why we have significance levels in statistics.