Mean difference in metaanalysis stata13

[Anastasia Tsiogka]

Hello statalist team!! I am using Sata13 for a metaanalysis. I use the command metan N1 varname1 sd1 N2 varname2 sd2. In the forest plot the results do not show the mean difference of my two groups. How can I correct that?
+----------------------------------------------------------------------------------------+
Study year UTDSAE~s BCVAlo.. DSAEKsd DMEKeyes BCVAlo.. DMEKsd
----------------------------------------------------------------------------------------
1. Chamberlain 2018 25 .16 .18 25 .04 .12
2. Mencucci 2020 18 .1 .04 18 .07 .07
3 Dunker 2020 25 .15 .11 29 .08 .14
4. Torras-Sanvicens 2021 10 . . 10 . .
5. Matsou 2021 28 .11 .09 28 .04 .13
+----------------------------------------------------------------------------------------+

In addition this variable cannot take price over 1. And at the forest plot it reaches 1.36. Thank you.

[David Fisher]

Hi Anastasia,

Firstly, to answer your question: as shown on the plot and on-screen, the summary measure is SMD = Standardised Mean Difference. This assumes that measurements in different studies used different scales, and have therefore been scaled by their standard deviations before pooling. This is why the result does not seem to be as you expect. SMD (specifically, Cohen's d) is the default in metan and in many other meta-analysis packages. However, to obtain the result you expected, you can add the option "wmd" to the end of the command; see below.


Secondly: thankyou very much for showing your data! Unfortunately, in the way it's presented the varnames are chopped and unreadable, so I have re-named them simply to "n1 mean1 sd1 n2 mean2 sd2":

* Example generated by -dataex-. For more info, type help dataex
. clear
. input str16 study int year byte n1 float(mean1 sd1) byte n2 float(mean2 sd2)
  "Chamberlain"      2018 25 .16 .18 25 .04 .12
  "Mencucci"         2020 18  .1 .04 18 .07 .07
  "Dunker"           2020 25 .15 .11 29 .08 .14
  "Torras-Sanvicens" 2021 10   .   . 10   .   .
  "Matsou"           2021 28 .11 .09 28 .04 .13
. end

. metan n1 mean1 sd1 n2 mean2 sd2, label(namevar=study, yearvar=year) wmd

Next: From the look of that plot, you are using an older version of metan. That's not necessarily a problem, because the syntax has not changed and neither will your results; but the plots now look a bit better and there are more options available. You can install the latest version (currently v4.06 12oct2022) by typing:

ssc install metan

(You may have to uninstall your existing version first.)


Best wishes,
David.