Hello, 

Q1. I'm running metaregressions within a single study (as opposed to a case control study comparing 2 groups - this is in the AES tutorial). The output images for the metaregressions are '_0' (differences between group 1 with minimum variable value and group 2); '_1' (differences between group 1 with maximum variable value and group 2), and '_1m0' (differences between group 1 with maximum variable value and group 1 with minimum variable value). How do I interpret these maps (_1,_0,_1m0) when only one group is being examined, as opposed to two?

At the moment, using just one group, I am interpreting _1m0 is the regions more active during higher value of the variable, vs lower values of the variable (within the same group) - is this correct? 

Q2. Can I run metaregressions on a categorrical variable? (I have the values in the sdm table as 1's and 0's to indicate the presence/absence of a categorical variable, and then I add that column as a regressor). Is there a better way to test the impact of categorical variables? 

Any help/input from anyone would be greatly appreciated - including references to relevant materials so that I ca read more up on this! 
Thanks. 

Best, 
Molly

Dear Molly, 

Q1. When you are performing a meta-regression for one group the results _1 correspond to the minimum value of the group. _1 correspond to the maximum value of the group, and _1m0 would be the slope, the difference between the minimum and maximum. So you are right, in your interpretation.

Q2. Yes indeeed you can perform meta-regression with categorical variables. It is a good way to analyze it.

Kind regards, 
Lydia

Thank you Lyda! 

That is really helpful; I had been trying to perform the differences analysis with two separate meta-analysis results (i.e. not combining them into one).
I have now combined them as you say, but the resulting '*_1m0_z_p.nii' has no activations. What's strange is that one a threshold it, the resulting '*_1m0_z_p_0.00500_1.000_10' map shows multiple widespread activations - many of which ook almost identical to the conjunction analysis (so that the differences map looks almost the same as the convergence map...). Are you not supposed to threshold this map? Perhaps it is already thresholded if it draws from a thresholded main analysis map? 

Thanks for your help
All the best, 
Molly

Dear Molly, 

Did you set a categorical variable to identify meta-analysis 1 (0) and 2 (1), and perform the linear model with that variable?

Best regards,
Lydia

Hi Lydia, 

I reran the analysis and it is working fine now! Thank you for your help. I have another related question :

I wish to look at the differences in fMRI activation between two fMRI meta-analyses. I have combined all the maps into one table, setting the maps from one (group A) to 0 and the other (group B) to 1. After preprocessing the data and conducting the mean analysis etc., I run a linear model - compare 2 groups. As the Second group indicator, the variable that is set to either 0 or 1 (denoting whether it is group A or B) is selected. This creates a map, e.g. groupAgroupB_1m0_z.nii.gz... does this need to be thresholded again? 

More importantly, is this the correct way to look at differences in activation? Group A completes one cognitive task, and group B another cognitive task, and I want to see the differences in fMRI activation between them (what is more/less active for group A than group B). Is what I have done the correct way of going about this? 

If so, I wanted to double check that I am interpreting the resulting maps correctly. If Group A is set to 0, and group B is set to 1, then do the activations in the "*_1m0_z.nii.gz" map show regions that are more active in group A, and less active in group B; or less active in group A and more active in group B? Or something else entirely... 

Your help is very appreciated! 

P.S. I am using the older AES-SDM v5.142 

Best, 
Molly

Dear Molly, 

Indeed you conudcted the proper analysis the investigate the differences between two different fMRI tasks. The resulting maps need to be thresholded again to check which voxels/clusters survived the defined threshold, in order to report your results. 

The results of the activation in **_1m0_z.nii.gz corresponds to a higher activation in group B with respect to group A. 

Best regards, 
Lydia

Originally posted by Lydia Fortea:
Hi, 

I just wanted to quickly clarify: did you mean "results _0 correspond to the minimum value of the group", as opposed to "_1"? 

Thanks! 

Hi Molly, 

yes! sorry for the confusion, 

Best, 
Lydia