I am conducting a meta-analysis on 16 studies. The outcome of interest in each study is represented by a proportion/percentage. I need to do meta-analysis of these proportions in STATA. So my question is how I am going to do meta-analysis of proportions in STATA?
Hi Tegene,
I did my meta-analyses via RevMan software. I am sorry that I couldn't send you the syntax.
Cheers,
Dear Hailay,
Is there a way to do meta analysis of proportions or prevalences by RevMan?
Thank you!
Article Metaprop: a Stata command to perform meta-analysis of binomial data
Better to use Metaprop command in STATA
there is also metaprop command in R
In Stata you can use metaan (ssc install metaan) and look at the help file (help metaan). There is an option for meta-analysis of proportions, while including all advanced meta-analysis options and graphs (unlike metaprop)
Thank you all! It was helpful! it is appreciated If you have additional suggestions!
I'm finding metaprop to be very useful for conducting meta-analyses of proportions. The article by Nyaga et al. (2014) linked by someone further above overviews the command and its rational. One key feature of this command is that it incorporates algorithms to address a common issue when conducting meta-analyses with proportions: confidence interval values being generated that are inadmissable, or outside the 0-to-1 range. The article also has sample code that shows how to customize the forest plot output. Very useful!
Chandrashekhar, in terms of your specific question, I think you are right. It looks like the numerator and denomator are the only variables needed to run the metaprop command. To generate a detailed forest plot, it is also common to have identifiers for studies, or in your case, WHO regions or World Bank categories. For example, when conducting a meta-analysis across authored studies, it is common to include in the dataset a variable for author and another for year. Then, these variables can be incoporated into the forest plot output by expanding on the metaprop code. Hope this helps!
Dear all
Do you have sample of forest plot done by SATA? Thank you all!!
Hi Tegene,
The metaprop Stata command can help you do the meta-analysis for proportions, including forestplots. For the detail, refer
Article Metaprop: a Stata command to perform meta-analysis of binomial data
I hope this helps,
Doyo
This is a great discussion. I am doing a meta-analysis of proportions from studies in east African countries. Metaprop is a useful command in stata for the analysis, it does everything without excluding studies that falls out 0 and 1 range.
@tegene
You can use metaprop command. However, you need to stabilize the variance using double arc-sine transformation before pooling.
Hello everyone.. I am trying to do a meta analysis of studies which report Odds Ratio (OR). I also have some studies which have reported only "proportions" and not OR.
My question is - can i convert proportions to OR using 2 by 2 table? And then use it along with the other OR data for forest plot?
Hi
But metaprop dont runn the command of metatrim ,so we can not do sensivity analysis .if begg or egger is significant. I will be appreciate if you help me to find the command.thank you
In the manuscript below you can find the procedures to run the metaprop command on stata.
https://archpublichealth.biomedcentral.com/articles/10.1186/2049-3258-72-39
@Bhavadharini Balaji, yes you ,ay convert the proportions into odds ratio as required for analysis. In fact if few studies report OR with reference to cat 1 and other studies report OR with category 2, then also you may take the reciprocal of one and hence obtained OR can be used.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis