admetan- p-value for pooled estimate?

[Natalie Malek]

Hello,

I have successfully use admetan to create a forest plot but after I updated to the newer version of the command, the p-value for the pooled estimate is no longer displayed. I only see I-squared. Can you tell me how I can display the p-value?

Thank you,
Natalie

[Marcos Almeida]

This thread may be helpful to you. It also refers to the explanation in the help files. Just type - help admetan - for that matter.

[Natalie Malek]

Thank you. I had already tried those options. The p-value was displaying with admetan in the older version I was using but once I updated it, it stopped showing. Nothing about this in the help file as far as I could tell.

[David Fisher]

Dear Natalie,

Sorry for not replying sooner! I am the author of admetan.

When you say "the p-value for the pooled estimate is no longer displayed", do you mean a p-value appearing to the left of the plot, in brackets, adjacent to the I-squared statistic? If so, that p-value was associated with the test for heterogeneity (i.e. I-squared), not for the pooled effect. In fact, this potential for confusion was precisely the reason that I removed this p-value from the default display :-)

If you wish to display the heterogeneity p-value in your plot, you can use the option hetstat(p). This should be documented in the help file.

There is currently no option to display a p-value for pooled effect, and in fact has never been (nor is there in the older command metan). To include this p-value (I would suggest on the right-hand side of the plot, next to or underneath the effect estimates), you would need to edit the forest plot in a similar way to the thread linked to by Marcos. I can help with this if you like.

Best wishes,

David.

[Ken Shores]

Hi David - Thanks very much for your work programming admetan. It's a great package. I was wondering if you could help me with something. I am interested in whether the critical values for the Cochran's Q heterogeneity test are sensitive to individual studies. The influence option plots the meta-analytic averages, dropping one study at a time. What I would like to do is loop over the individual studies, drop one of them, and then obtain both the Cochran's Q statistic and its associated p-value.

When I look at the return list, however, it looks like r(obstats) contains the p-value for the overall effect but not the p-value for the test of heterogeneity. I would like to store the p-value for the test of heterogeneity and look at those later.

Thanks very much in advance for any help you might provide.
-Ken

[David Fisher]

Hi Ken,

Sorry for not replying sooner -- I hope this will still be useful to you!

If you use the saving() option together with influence, you should find that the saved dataset contains the Q and associated p-value corresponding to each analysis with the labelled study omitted/dropped, under the variable names _Q and _Qdf.

Somewhat ironically, the reason for my recent absence from StataList was that I have been working on an update to the main user-written package metan, which was finally released earlier this week . I had intended this update to incorporate all the functionality of admetan. However, when looking at your particular question, it turns out that the saving() option in metan does not seem to save the variables _Q and _Qdf. I'm not quite sure why that should be the case, and I shall fix it in the next metan update. In the meantime, hopefully admetan will give you what you need.

Best wishes,

David.