Stata, a statistical software package by StataCorp, boasts an impressive suite of meta-analysis features. metan is the Stata module for fixed and random effects meta-analysis. metafunnel “plots funnel plots: graphical displays used to examine whether the results of a meta-analysis may have been affected by publication or other types of bias.” These two commands did not come with Stata 10 by default; users of this and a number of subsequent versions of Stata have to download and install these user-written meta-analysis commands themselves.
Below is a summary of how I installed metan and metafunnel in Stata 10. Depending on your version of Stata, YMMV.
I essentially used the guide available in Chapter 18 of Systematic Reviews in Health Care: Meta-Analysis in Context, 2nd Edition “Meta-analysis in Stata”, by Sterne, Bradburn, and Egger, but with some small changes and additions which I will outline below.
Step 1: Update your installation of Stata.
. update all
Step 2: Install the user-written commands, e.g. metan, metafunnel.
I installed the commands from within Stata, i.e.
Help → SJ and User-written Programs → STB
You will see a list of commands associated with Stata Technical Bulletins (stb). I simply followed the instructions found in Sterne et al.’s guide:
I found that, in my installation of Stata, the packages had an underscore, e.g. sbe24_1, instead of a decimal point, e.g. sbe24.1.
In addition, I installed the following metafunnel command within Stata by typing:
. ssc install metafunnel
This command enables creation of funnel plots with pseudo 95% confidence limits.
Step 3: Check for updated versions or new commands:
. update all
. search metan
Having performed the steps above, I was unable to use the metan command to perform a fixed and random-effects meta-analysis. The problem was an old version of metan, specifically metan 1.0! You can check your version of metan:
. which metan
To update the version of metan, simply enter the following command on the Stata command line:
. ssc install metaaggr, all replace
Completing this last step enabled me to conduct a fixed and random-effects meta-analysis.
The steps above should not take longer than 5 minutes. Enjoy!