Hello!

I recently conducted a linear model to formally compare the effect of task type. My contrast (Cognitive>Physical) was set up so that studies that used  cognitive effort were coded as 1 and studies with physical effort were coded as 0. I was a bit confused by the results and was wondering if you could help me with the interpretation.

So far, every other meta-regression I've run has both positive and negative SDM Z values, so I assumed that positive SDM Z values suggests regions that are associated with contrast=1 (Cog>Phys) and that clusters with negative SDM Z values are associated with contrast=0 (Phys>Cog). With the linear model described above, however, I only have negative SDM Z values. It seems unlikely that there are no activations specific to cognitive effort and only activations associated with physical effort, so I'm wondering if I've been interpreting these results incorrectly.

For example, does a deactivation in this linear model (Cognitive>Physical) suggest that these regions that are deactivated during cognitive effort, activated by physical effort, or both?

I tried conducting a subgroup analysis to clarify a bit. My subgroup analysis only yielded positive SDM Z values (activations) for physical effort and negative SDM Z (deactivations) for cognitive effort, so this seems to confirm my last interpretation. Interestingly, I found regions in the subgroup that weren't captured in the linear model. What does this mean?

Also, I looked at funnel plots of the linear model and they are symmetrical, but shifted in the negative direction. It seems like they are skewed based in the direction of the effect. Should I just work with the publication bias funnel plots for the subgroup analysis?

Finally - last question - my excess significance test never works because I have a few NA values in my t_thr column. The SDM tutorial had suggested we put NA if we genuinely don't know and that SDM will fill it in to the best of it's ability. How do I get around this to conduct the excess significance test?

Thanks for answering all these questions. I really appreciate any help!

Best,
P

Dear Dr. Lopez-Gamundi

Z>0 would mean cognitive > physical, which may include several possibilities: activation in cognitive greater than activation in physical, only activation in cognitive, only deactivation in physical, deactivation in physical greater than deactivation in cognitive...

Z<0 would mean physical > cognitive, which again may include several possibilities: activation in physical greater than activation in cognitive, only activation in physical, only deactivation in cognitive, deactivation in cognitive greater than deactivation in physical....


It is very likely that the results of this model and the subgroup analyses have differences because their statistical power is different.

Funnel plots are usually conducted with the mean analysis. For meta-regressions, their interpretation would be more complex.

The excess significance test cannot be conducted if the thresholds of statistical significance are unknown.

Hope this helps,

Joaquim