Forest plot

[Soodabeh Navadeh]

I am doing a meta-analysis in which I will need to generate pooled prevalence rates. Since my data is not normally distributed I transformed p, L95 and U95 by logit(p), logit(L95), logit(U95) and did the meta-analysis using metan command. The problem is that in the forest plot the result is shown in the logarithmic scale that does not make sense for clinician. Is it any option to back transforming the result and have the forest plot with prevalence and percents? Thanks, Soodabeh Navadeh MD MPH Research Specialist University of California, San Francisco

[David Fisher]

I'm not sure about percentages (if you mean having the "%" symbol appear in the graph, that may need some manual tinkering), but you can easily back-transform using the eform option.

[Soodabeh Navadeh]

Thanks David. It helps.
eform option calculates the exponential but I wanna back-transform from logit. It means I need to use Inverse logit (invlogti). Do you have any suggestion?

[David Fisher]

Ah, of course, sorry!

OK, as far as I'm aware metan does not support logit back-transformation. Furthermore, on doing a bit of searching, it seems that meta-analyses of prevalences/proportions is generally a bit trickier than usual meta-analysis. Here is a previous Statalist post that in turn links to some other posts on the subject: http://www.statalist.org/forums/foru...on-proportions

There is also a user-written command metaprop which is designed for the meta-analysis of prevalences/proportions, which may be helpful, although I have never used it myself.

Thanks,

David.

[Soodabeh Navadeh]

Thanks again David
metaprop can solve my problem Just one problem remains:
I found in the help that "by" is defined for metaprop. But I encounter "invalid syntax" error when I use it.
Do you have any idea, please?

Thanks,
Soodabeh

PS: This is a very useful paper about metaprop:
http://www.ncbi.nlm.nih.gov/pubmed/25810908

[ben earnhart]

paste your *exact* command and output, preferably in

code

tags (see FAQs), and it might be possible to decipher your "invalid syntax" error.

[Soodabeh Navadeh]

Hi Ben,
This is the command and output:

. metaprop case_hiv total_hiv , random by(inject3)
invalid syntax
r(198);

[nora hendriyeni]

Hi everyone,
I am currently running my STATA and after trying so many times ..I keep having this message:
invalid syntax r(198)
What is this? How to fix it?
Please help me, I know nothing about programming, coding etc. I just follow the steps that my friend's used, and it work with her. we are calculating the same variable.

[nora hendriyeni]

this is my command:

gen residuals=.
forvalues years=2000/2012 {
2. forvalues types=1/2/3/4/5/6/7 {
3. reg tass drev ppe if years==`years' & types==`types'
4. predict temp, residuals
5. replace residuals= temp if years==`years' & types==`types'
6. drop temp
7.}
8.}

Once I enter the no 8.} I got this message:
invalid syntax r(198)
Please help me...I am running out of time (submission deadline).

Thanks

[Rich Goldstein]

the syntax for "2." is wrong - you probably want "foreach" here - see the help file as the actual syntax will depend on what "1/2/3/4/5/6/7 " actually - if they are numbers within years you can probably use

forval types = 1/7 {

be careful here as this will give you 7*13 = 91 regressions - is that what you want?