metafunnel: funnel plot for mean differences

[Talia Sherrard]

Hello,

I am conducting a systematic review and the majority of studies present their results as means and SDs of groups, with a p-value generated from chi squared/t test/ANOVA.

I have conducted a meta-analysis of these effect estimates using the metan command:

metan n1 mean1 sd1 n0 mean0 sd0

I would like to add a funnel plot to my paper, and had been planning on using the metafunnel command to do this. The user manual says it uses the same syntax as metan, however it does not seem function when attempting to input mean differences. The code below returns the error code "too many variables".

metafunnel n1 mean1 sd1 n0 mean0 sd0

What would be the best way of generating a funnel plot for studies reporting mean differences? If anyone with more experience in this area could advise it would be hugely appreciated. Many thanks

[Marcos Almeida]

The help file of metafunnel indicates the code is something like: metafunnel meandiff semeandiff

[Suhail Doi]

Probably you will need to drop metan and use admetan

ssc install admetan

admetan n_treat mean_treat sd_treat n_ctrl mean_ctrl sd_ctrl, ivhet wmd

(note that the ivhet model is in the code and is more reliable than RE models)
The admetan module adds _ES and _seES to the variable list
Then you can use the lfk module

ssc install lfk

to create a Doi plot and LFK index as follows:

lfk _ES _seES

These are more reliable then funnel plots and p value driven tests

Best
Suhail Doi