Meta-analysis of means

[Martin Mueller]

Dear all

I have a bunch of different studies with measurement of a continous outcome presented by mean and se.
I like to pool the means of each study, to get to a pooled mean.

Unfortunately, Stata defines a meta-analysis as pooling of effect sizes only (in contrast to wikipedia :-)) - and so is the - meta - command structured.
Googling the issue, I found that there is a - metamean - function in R, but I wonder if there is an analogy in Stata (as to proportions the user written - metaprop - command), e.g. the metan function.

My data and naive guess would be the following.

clear all

input mean_s se_s
14 4.58
13 2.58
15 1.58
16 2.86
16 1.61
end

metan mean se, random

Is this correct or is there a better/different solution?

Best wishes & thanks!

Martin

[Leonardo Guizzetti]

Effect size is a generic term in statistics, and refers to any parameter estimate. Common choices in meta-analysis are measures of treatment efficacy because the scientific questions are focused on these: odds ratios, risk ratios, differences in means or proportions, etc. However you are not limited to those quantities. It is occasionally of interest to pool single means, proportions or correlations, as is your case. When using means as the effect size, two options are usually presented, the standardized mean or raw mean (and their differences). Standardizing is useful if the means are measured on different scales to put them on a common scale, such as measuring body weight in kilograms and pounds. Raw means are just that, no transformations applied, and useful when emerging is measured in the same scale.

For your case, you already have the means and standard errors, so you can do this, for example.

meta set mean se, random
meta summarize

For more information see -help meta set- especially example one.

[Martin Mueller]

Thank you for the perfect and super quick help, Leonardo! Works perfectly.

And to my delight, your solution gives the same results as the one with the - metan - command (at least with my data).