External Email - Use Caution

Hi Julian,

Am 11.03.2022 um 11:06 schrieb Julian Macoveanu <julian.macoveanu@gmail.commailto:julian.macoveanu@gmail.com>:

External Email - Use Caution

Dear Kersten,

Thank you for your answer, I have some follow-up questions to the first two points before I can move on.

1) Regarding family-relation, since three random effects may not work, did you mean that a simpler model with 2 random effects (intercept and family ID, thus leaving out slope which I assume is time from baseline) would work? While forgetting family ID is an option, it would be a model that may attract criticism.

I generally like the idea of proceeding from simple to more complex models, assuming / checking that results do not fundamentally change when models get a bit more complex. In your case, I initially considered looking at a sequence like 1. random intercept (subject), 2. random intercept (subject) and random slope (time), 3. random intercept (subject), slope (time), and another, higher-level random intercept (family). But you are right, the two random intercepts (subject ID and family ID) model would also be worth checking.

In my previous response I was a bit skeptical if adding a higher-level random effect (ie. family ID) would work. This was because I remembered other software (e.g. R’s lmer package) that require a specialized syntax for nested vs. crossed random effects, and I did not immediately see a way how to implement in FreeSurfer’s lme toolbox.

However, I now think that in the end it boils down to simply checking the covariance matrix of the random effects (that should be stats.Dhat). If the program is able to construct this matrix so that it correctly captures the dependencies between the observations, then I see no reason not to proceed with a higher-level random effect. The disclaimer is that I have not tried this before, and that the above just reflects my current understanding.

2) Regarding multiple comparison correction, I understand how to perform the procedure you suggested (second option) to simply calculate the corrected p threshold using pcor = -log10(pth) and then use pcor as threshold for the sig.mgh. However after running FDR2: [detvtx,sided_pval,pth] = lme_mass_FDR2(F_lhstats.pval,F_lhstats.sgn,lhcortex,0.05,0);

While detvtx is not empty (a few vertexes survived) and the pth value is calculated to 1.00000e-30 and therefore pcor to 30. This doesn’t seem right. What is also strange is that I saw a few times a more reasonable pth in the range of e-5 using the exact same data and procedure. I cannot replicate this in the same analysis in R2021, now I only get the e-30 (I only tried left cortex thickness so far).

It’s a bit hard to tell if something went wrong without looking at the data, but here are two suggestions: one would be to look at the beta weights for that contrast, which should show a somewhat smooth and somewhat anatomically plausible pattern. The other would be to look at other contrasts such as the simple effect of time, or the group differences at baseline, and also check their spatial pattern and their magnitude. Both might help to judge whether or not something went wrong with the analysis.

Also, you mentioned before that you had issues (and possibly different results?) with Matlab’s R2019 release, in particular with its parallel processing tools; if you have a way to share the error messages, that could be another way to proceed.

Best regards,

Kersten

Best, Julian

_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://secure-web.cisco.com/1nqTgGbepYhNcklULiiDATEn_lUePeuAHYnJTeTZYwfvyyH... The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Mass General Brigham Compliance HelpLine at https://secure-web.cisco.com/1nVHzpqdExBMekrwnonIULEBvQgR6xo2X_LJ_706lTfiiDG... https://secure-web.cisco.com/1nVHzpqdExBMekrwnonIULEBvQgR6xo2X_LJ_706lTfiiDGoKJb9D2_RjDRq6aqGSniJiFVwKp-Sgm0fqwiD9btK4UEgBY8sQ001F2RF9JWA_WWby1EZ_vFZoWCRbbXumiMFoOuqymfkzw4IWJ-mbPMH3Sa8PKnxYGe5-nxiJgMrr8jKdquD3EKxY7BuTSlAmLJWqJwLSnGrtjr9bzqERJtVRSuz1_eLsq7fZINp_VyRgO7OinbNYcoyP8rS3clsMDpzLluxgKRmNV5VHT_e0shj3v82CeydEIIYugGlKxFPZhrt0yADQFVL3N0t2Znek/https%3A%2F%2Fwww.massgeneralbrigham.org%2Fcomplianceline . Please note that this e-mail is not secure (encrypted). If you do not wish to continue communication over unencrypted e-mail, please notify the sender of this message immediately. Continuing to send or respond to e-mail after receiving this message means you understand and accept this risk and wish to continue to communicate over unencrypted e-mail.

-- Kersten Diers Image Analysis Group (AG Reuter) German Center for Neurodegenerative Diseases (DZNE) B.1.114 | Building 99 | Venusberg-Campus 1 | 53127 Bonn | Germany https://secure-web.cisco.com/10j6bPcc9yA1LCGZmQRAyb6Mcc1ZFfEJG2c7d1EetFmlsDV... Phone: +49 / 228 / 43302 - 381