External Email - Use Caution Hi Donatas,
please find my thoughts in-line. Best regards, Kersten On So, 2018-05-27 at 11:59 +0200, Donatas Sederevicius wrote: > External Email - Use Caution > > Dear Freesurfer experts, > > I have some doubts while running LME analysis on longitudinal data. > The main goal is to check whether longitudinal BMI scores have any > impact on the longitudinal cortical thickness changes. The model I’m > thinking of is as follows: > > > Y_ij = b0 + b1*time_ij + b2*BMI_ij + b3*Skyra_ij + b4*Prisma_ij + > b5*gender_i + b6*bslAge_i. > > > And here is an example design matrix X for the model above: > > > ------------------------------------------ > > ones|time|BMI |Skyra|Prisma|gender|bslAge| > > ------------------------------------------ > > 1 |0 |0.62|0 |0 |0 |1.89 | > > 1 |3.2 |1.56|0 |0 |0 |1.89 | > > 1 |7.2 |2.04|1 |0 |0 |1.89 | > > ------------------------------------------ > > 1 |0 |1.1 |0 |0 |1 |1.67 | > 1 |1.5 |0.9 |0 |0 |1 |1.67 | > 1 |4 |0.9 |1 |0 |1 |1.67 | > 1 |5.3 |1.3 |0 |1 |1 |1.67 | > > ------------------------------------------ > > 1 |0 |0.7 |0 |0 |0 |0.89 | > 1 |1.2 |0.5 |0 |0 |0 |0.89 | > > ------------------------------------------ > > > Note that BMI and bslAge are z-scored, and there are 3 different > scanners: Avanto (base), Skyra and Prisma. I’m trying to account for > differences between scanners as well. For some participants, the > first two timepoints were scanned with one scanner and the last time > point with a different one, so it varies within the subject. Do I > account for the scanner differences in a correct way? Should I add > interactions with the time variable, as time*Skyra? Including scanners as you currently do is fine in my eyes. I would probably also check whether or not results remain stable when observations after a within-participant scanner change are excluded from analysis (in order to have the same scanner for the same participant at all included timepoints). > Another question. Since BMI is longitudinal, time-variant, should I > add an interaction with time as well? In my eyes, if you want to answer the question whether or not the effect of BMI scores on thickness depends on time, or (vice versa) whether or not temporal changes in cortical thickness depend on BMI: then yes, you should add an interaction term of BMI and time - and also construct and test a corresponding contrast. > > Assuming that the above design matrix is “correct”, I would use the > following contrast [0 0 1 0 0 0 0] to answer my question whether > longitudinal BMI scores have any impact on longitudinal cortical > thickness changes, right? This contrast will, for your proposed design matrix, test for a positive relation of BMI and cortical thickness, irrespective of time. So this is, if I understood correctly, probably not what you want. Rather test an interaction term, see above. > > P.S. I’m running spatiotemporal models with one random effect - > intercept. > > > Thank you for the answer! > > > Best, > > Donatas > > > > > > > _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
