Dear experts,

Given that the current SDM software does not allow complex regressions (e.g., non-linear regression, see this post), I am thinking to extract data and apply the meta-regression in R. However, I encountered several questions.

(1) Is it acceptable to extract the results based on the imputations of the simplest analysis, such as a Mean analysis, and then apply other analysis manually on the extracted data?

(2) To conduct other analysis, I probably need to extract the results for each voxel. An exhaustive way could be creating a mask for each voxel, which is apparently infeasible. Is there an easy way to achieve this?

(3) I wonder if the .sdmimage.bin files in the mi/i00** folder could be read into R or Matlab, or if they could be converted into nifti files? I found a function sdmimage2nii.exe, which seems to do this work, but when I ran the code it simply displays (without reporting error) but did not create the MyFile.nii.gz file at all.

Do you have any suggestions? Thanks in advance.

Best,
Guochun

Hi experts,

Sorry to bother again. These same questions are still haunting me. Could you please give me any suggestions?

Thanks a lot!

Guochun

Dear Gouchun, 

Sorry for the late response.

1) If possible, we would recommend specifying the meta-regression in SDM, e.g., using the regressors x and x^2 for a quadratic regression; but if you really want to perform other type of regressions you are free to do them.

2) As you asked in the point 3, you can conver the sdmimga.bin into nifti and then use those images for further analysis

3) in the third parameter of the function, you should add the number of voxeles in the image depending of the size of your voxel (2mm: 193768; 4mm: 24294; 8mm: 3038)

Please not that the extracted valoes of the mean analysis are safe for the mean analysis but may be slightly biased for other analyses, because the imputation calculations conduct an internal leave-one-out analysis to avoid biases but only in the direction of the analysis conducted.

Best,
Lydia

This helps a lot. Thanks!