1
$\begingroup$

I am conducting a meta-analysis of single-group studies that result in proportions/percentages of correct categorisations of someone's gender. The goal is to see whether these studies correctly guess someone's gender above chance (50%).

As they are single groups, can I conduct a meta-analysis by adding a 'control group' to each study that has the same total sample as the study group with the baseline result of 50%? Or is there a more sophisticated way of conducting a meta-analysis of these single-group studies?

$\endgroup$
  • $\begingroup$ You say you want to see if they "correctly guess", are you trying to assess the classification performance of a fitted model? What were the studies? Do you have the actual data from the studies or just summary statistics? $\endgroup$ – gung Sep 12 '16 at 11:59
  • $\begingroup$ Perhaps an example of a study will help clarify things. Participants were involved in computer mediated communication where they either posed as someone from the same gender (male posing as male/female posing as female) or the opposite gender (male posing as female/female posing as male). At the end of the conversation, participants were asked whether the other person is male or female. So I have percentages of the number of right guesses at the end of the conversation as well as the sample size. $\endgroup$ – Tommy Sep 13 '16 at 9:09
  • 1
    $\begingroup$ I see no reason for making up data and pretending as if you had two groups in each study. You can just meta-analyze the percentages, or rather, the proportions directly. $\endgroup$ – Wolfgang Sep 13 '16 at 21:32
  • $\begingroup$ Ah yes, meta-analysis of proportions, that is what I was looking for. Thank you for getting me on the right track Wolfgang. $\endgroup$ – Tommy Sep 14 '16 at 16:09
1
$\begingroup$

There have been a number of answers on this site which may be helpful

Using variance from a Jeffreys prior in a Prevalence meta-analysis

Methods reference for meta-analysis of observational prevalence studies

and if you are using R @Wolfgang has a worked example on his website here

$\endgroup$
  • $\begingroup$ Thank you for your helpful answer. (I up-voted it but as my reputation does not (yet) exceed 15, it doesn't do anything.) $\endgroup$ – Tommy Sep 22 '16 at 10:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.