Dear SDM Team,
I am a psychology student and currently conducting a meta-analysis with the „ES-SDM"- program for my master thesis.
I have a short question about the interpretation of the Q-statistics.
I know that the Q-statistic for the mean analysis represents inter study heterogeneity for brain regions across all studies, where the Q-statistic for linear models or regressions with more than two groups or predictors represents an overall effect across the groups or predictors.
In my linear models and regression I have only two groups and one predictor so I was wondering, if I can still interpret the Q Statistic in the same way as if there were more groups/predictors?
Many thanks in advance.
Best regards,
Mareike

Dear Mareike,

I'm not sure whether I understood the question. As you say, SDM returns two Q statistics. One is a statistic for heterogeneity, and SDM calls it "QH". The other is a statistic for the overall effect of the first 2 regressors of a linear model, and SDM calls it "Q2". SDM estimates "QH" in any linear model (including the mean, which is indeed a very basic linear model with no regressors), whereas it only estimates "Q2" when there are two or more regressors.

Hope this helps,

Joaquim

Dear Joaquim,

Thank you for your answer!
I was not aware that the QH statistic is always the statistic for the heterogeneity and the Q2 the statistic for the overall effect.
I thought that in the comparison of two groups e.g. patient vs. control, the QH represent the overall effect, which I now know is not the case.
So thank you very much for clarifying the Q-statistics for me.
Kind regards,
Mareike

Dear SDM Team,

I have another question about the interpretation of the QH Statistics.
I already know that the QH statistics estimates the heterogeneity for any linear model.

So in the simplest model, the mean, it accounts for heterogeneous brain region across all studies.

But if I calculate a simple linear model with one dichotomous regressor (in my case I divided the studies in two task groups), what exactly tells me the QH statistics now?

Many Thanks in advance,

Mareike

Dear Mareike,

QH statistic reflects the heterogeneity across studies. Let's imagine that in a meta-comparison, for example, a given voxel has values around 5 in studies of group A and values around 8 in studies of group B. Let's imagine, also, that in the absence of heterogeneity values in group A are in the 4-6 range and values in group B in the 7-9 range (there's some variability due to sampling error). However, in the presence of heterogeneity, values in group A could be e.g. in the 1-9 range and values in group B in the 4-12 range.

Hope this helps,

Joaquim

Dear Joaquim,

Thank you for your fast reply and the little example!
It really helped to clarify the meaning behind the QH-statistic.
Thanks again!

Best regards,
Mareike