1
$\begingroup$

I am doing a systematic review on the risk factors that predict infection after liver and pancreatic surgery. These risk factors include parameters like age, sex, smoking, pre-operative chemo etc.

I have a number of studies but not all report their findings in the same way, so I am having trouble doing the meta-analysis. Some of the studies only report a p-value. Beyond simply counting the number of studies that have significant P-values for each risk factor, is there another way to derive an overall p-value for each risk factor?

$\endgroup$
4
  • 1
    $\begingroup$ Do you know the direction of the effect? $\endgroup$ Feb 10 '20 at 17:06
  • $\begingroup$ You can work back from a p-value to a standardized difference and variance, if you know the sample size (and the direction). $\endgroup$ Feb 10 '20 at 17:07
  • $\begingroup$ I do know the direction of the effect and sample size. $\endgroup$
    – Keno
    Feb 11 '20 at 11:13
  • $\begingroup$ Here is a worked example using R: metafor-project.org/doku.php/tips:assembling_data_smd $\endgroup$
    – Wolfgang
    Feb 11 '20 at 18:51
4
$\begingroup$

There are many ways to calculate an overall $p$-value: Edgington's method, Fisher's, Lancaster's, Stouffer, Tippett's and several others. They all have their advantages and disadvantages. Since you have tagged this R you may be interested to read the vignettes for the metap package available from CRAN which contains some guidance on the choice. It is too long to include here.

Disclaimer: I am the author of that package. There are other packages available, see the CRAN Task View on MetaAnalysis for details (I maintain that too).

$\endgroup$
1
  • $\begingroup$ Thank you, will look into that package $\endgroup$
    – Keno
    Feb 11 '20 at 11:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.