4

I understand how to find the weighted mean. However, I am wondering if it is possible to find mean overall age (for all the studies) when I have a list of means and medians for the ages (e.g., some studies present age as a mean and other present it as a median... some simply present a range). Do I just take the mean of the studies that present age as a mean?

aspire94
  • 261
  • 1
  • 4
  • 1
    There is not unique solution to this problem. – Michael R. Chernick Mar 09 '17 at 00:38
  • 3
    If you are willing to assume that age has a symmetrical distribution then the median also estimates the mean and vice versa. However since you tag this meta-analysis I suspect you want to use the standard errors as weight and estimating them is going to be very tricky. There are some interesting suggestions and discussion in this Q&A http://stats.stackexchange.com/questions/88813/meta-analysis-of-means-and-medians-in-r?rq=1 – mdewey Mar 09 '17 at 10:22

1 Answers1

1

So this does not go unanswered I copy my comment here as an answer.

If you are willing to assume that age has a symmetrical distribution then the median also estimates the mean and vice versa. However since you tag this meta-analysis I suspect you want to use the standard errors as weight and estimating them is going to be very tricky. There are some interesting suggestions and discussion in this Q&A Meta-analysis of means and medians in R?

mdewey
  • 16,541
  • 22
  • 30
  • 57