Edit to palm_takeargs to correct an issue where multiple inputs would cause PALM to crash if one of the multiple inputs is a volume while another is a surface.

[jocwilliams]

This issue is when PALM was computing the adjacency matrix for spatial statistics. Even though these adjacency matrices are only necessary for computing spatial statistics using surfaces, PALM was attempting to calculate the adjacency matrix for volumes as well. This was solved by adding a check while iterating through inputs for whether or not the given input is a volume using plm.Yisvol(i).

[andersonwinkler]

Hi John,

Thank you. I'll review.

Could you tell me more about your use case? Normally we'd enter multiple inputs to correct across them, which is ok for t-stats, but these need no adjacency; for spatial statistics, that need adjacency, we can't correct across representations as these stats aren't comparable. For example, we can't correct cluster extent from voxel and surface representations as cluster sizes in each case follow different distributions.

But depending on the data, perhaps there could be exceptions. Let me know.

Thanks!

All the best,

Anderson

[jocwilliams]

Hi Anderson, Thanks for reviewing! And apologies for taking some time to reply. To be honest, the main use case was one of convenience and curiosity. I wanted to run PALM with all three components of a CIFTI (2x gii + nii), but using a single call to PALM and single output directory, by supplying them as three inputs. PALM was crashing with these three inputs even with corrmod disabled. When it didn't work out of the box, I wanted to know why, since in theory from the way PALM is set up documented it seemed that it should be possible. When I looked into it, it appeared to be simply because having the number of surfaces > 0 and choosing to do spatial statistics was causing all inputs to be subject to the calculation of surface areas and adjacency matrices, rather than just the surface(s). I was also curious about what the correction over modalities procedure would generate in this case, and had been meaning to ask you for your thoughts for quite some time. I am grateful for your explanation in your previous email. I do wish there were an alternative to the seemingly aggressive Sidak/Bonferroni correction across the three inputs, but perhaps I'll have to take up that issue with nature. By the way, I also wrote some code to generate and run PALM using an Excel spreadsheet. I was a bit hesitant to share since I do understand the logic behind mandating that users use scripts instead, especially for a tool with the capabilities of PALM, but I and others I have worked with have found it very helpful. I wanted to run it by you to make sure that you are okay with me sharing it more broadly. I just created a repo for it, though this code was written well over a year ago with some minor tweaks since. https://github.com/MMTI/Palm-From-Excel/tree/main It's not really designed for public consumption yet, but it's not too complicated to look through or use as-is. The main driver is Example_Main_Palm_From_Excel_Script.m, and the function that does the work is palm_from_excel.m. A spreadsheet template is in ExcelFiles/Excel_for_Palm.xlsx . Thanks again! Best, John From: A. M. Winkler ***@***.***> Sent: Sunday, July 2, 2023 11:46 PM To: andersonwinkler/PALM ***@***.***> Cc: Williams, John ***@***.***>; Author ***@***.***> Subject: Re: [andersonwinkler/PALM] Edit to palm_takeargs to correct an issue where multiple inputs would cause PALM to crash if one of the multiple inputs is a volume while another is a surface. (PR #45) CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe. Hi John, Thank you. I'll review. Could you tell me more about your use case? Normally we'd enter multiple inputs to correct across them, which is ok for t-stats, but these need no adjacency; for spatial statistics, that need adjacency, we can't correct across representations as these stats aren't comparable. For example, we can't correct cluster extent from voxel and surface representations as cluster sizes in each case follow different distributions. But depending on the data, perhaps there could be exceptions. Let me know. Thanks! All the best, Anderson - Reply to this email directly, view it on GitHub<#45 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AHOEGLPPRRYNV5LOW3C72J3XOI6A7ANCNFSM6AAAAAAZ3SRPA4>. You are receiving this because you authored the thread.Message ID: ***@***.******@***.***>> This e-mail message, including any attachments, is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by e-mail and destroy all copies of the original.