|
Dear all,
Currently I am conducting a meta-analyses and meta-regression and I am encountering the following problem. I have first done my mats-analyses in review-manager but also wanted to perform a meta-regression (with fairly simple co-variates), so I created a SPSS database. To do the meta-regression, I use the metareg.sps spss macro from David Wilson. When I am running a 'blank' model (without any of the co-variate included), using the fixed effects model I get a Q value that is the same as the Q value in review manager and is the same as the Q value when running the meanES macro (the meta-analyses data). In contrast, when I run the blank random effects model (methods of moments) I get very low Q values, indicating low heterogeneity which is not the case. The value is very different from what I get in review manager and also very different from the Q value for random effects in the meta-analyses itself.
Does anyone know why this is? It would really help me a lot!
Thanking you in advance!
Best wishes, Sanne
|