Hi all,

I was wondering whether it is possible to include multiple contrasts from the same study into one analysis? Can the program account for the dependence between these results?

Thank you very much in advance,
Nicky

Dear Nicky,

this will be indeed possible and very easy rather soon!

In the meantime, please contact me if you want me to send you a beta version that it has this feature already working.

With the best wishes,

Joaquim

Thank you very much! I will discuss this with my colleagues.

Best,
Nicky

Dear SDM Team,

I have a similar question about the inclusion of multiple contrasts from one study:

If a study examines 3 groups (e.g. patient group, high risk group and control group) in the same task and reports the contrasts for:

Patients > controls
High-risk > controls

Is it possible to include both contrasts in the meta-analysis even if both experimental groups were tested against the same control group?
Or should I strictly include only one of the contrasts because of dependent samples (same control group) in the two contrasts?

Thank you very much in advance!

Best regards,
Mareike

Dear Mareike,

I think that there are different ways to deal with this situation. For instance, if it only happens in one or few out of a lot of studies, I think you could include both contrasts in the meta-analysis as if they were independent, and understand that the analysis is "only slightly incorrect" and thus the p-values will be "only a bit less accurate". On the other hand, you could alternatively combine both contrasts in a single contrast, which is more statistically correct. Or you could model the specific dependencies, but this is difficult and anyway not implemented in SDM.

Hope this helps,

Joaquim

Dear Joaquim,

thank you so much for your fast reply!

It is the case for 3 studies and I have 25 without and 28 with the doubled contrasts. Counts that still as a few?

And how would I combine the contrasts from one study? Should I simply put the input files in one input file together and sum the N of the experimental groups?

Thank you very much in advance!

Best regards,
Mareike

Dear Mareike,

Do you mean that most studies have two contrasts to be included? If this is the case, I think they are not "few"...

To combine the contrasts of a sample into a single contrast you may use the "Combine images" button, soon available in SDM.

Best wishes,

Joaquim

Dear Joaquim,

I'm sorry for expressing myself before so unclear and making the question unnecessary complicated.

I mean that only 3 studies have 2 contrasts.
The rest (22 studies) have only 1 contrast to be included.

So I would like to know, if 3 studies still count as a few?

Hope you will help again!

Best regards,
Mareike

Oh, I see. I guess that it may be reasonable to think that they are too few, then. That said, if you are not in a hurry, you can still combine them using the next version of SDM.

Thanks for the good advice and the really quick feedback!

Best regards,
Mareike