Hi Anderson,
Thanks for pointing me in that direction!
Could you please see whether the objectives could be met using the the steps and design matrices outlined below?
1) 4D Imaging data Concatenated according to subject ID in ascending order (Subject 1 baseline ... subject 1 follow-up, Subject 2 baseline ... subject 2 follow-up, Subject 3 baseline ... subject 3 follow-up, etc).
2) Design matrix
- To capture the between - subject effect of age, I take the average age between the 2 time points: (Age at baseline + Age at baseline + Scan Interval) / 2
- For each subject, scan intervals were code as 0 and Years from Baseline (2.9 ... 3.1 ... 2.82 ... etc). Next, across the full sample, I demeaned the interval.
- To capture the pure within subject covariate: age_within = age at baseline and follow-up - the averaged age. This is the time variable that I am most interested in now
-
2) Abbreviated Design matrix (red represents negative values).
3) Contrast
Are there differences in brain volumes between both groups on average?1 0 0 0 -1 0 0 0 -1 0 0 0 1 0 0 0 Is brain atrophy over time more significant in Group 1 vs Group 2?
0 0 0 -1 0 0 0 1Is brain atrophy over time more significant in Group 2 vs Group 1?
0 0 0 1 0 0 0 -1Is there age-accelerated brain atrophy in Group 1 vs Group 2?0 -1 0 0 0 1 0 0
Anderson, would you be kind to show me PALM commands to run these tests? Based on the link you sent, I am still not sure how to include the within-block or whole block exchangeability into this model?
Lastly, sorry if this is a silly question, how will this be different compared to the modified Sandwich Estimator approach using the "SWE" command? In that model, I understand I will have to include additional information in the design.sub which looks like this.
Thanks ever so much for your guidance.
Regards,Shane
On Tuesday, February 11, 2020, 10:33:07 AM GMT, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Shane, Tom,
It should be possible to use randomise or PALM in this case. See the discussion in this thread: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;3a78265b.1512
All the best,
Anderson
On Thu, 6 Feb 2020 at 05:57, Shane Schofield <[log in to unmask]> wrote:
Hi Professor Thomas,
Thank you for introducing me to the modified SWE toolbox
Say, I am trying to see whether two groups differ in their rate of change in grey matter volumes over 2 ( mean follow up is about 3 years), and also ask whether this rate of change correlate with other variables.
The design.sub would look something like
| Order of Scan | Visit Number | Group |
| 1 | 1 | 1 |
| 1 | 2 | 1 |
| 2 | 1 | 1 |
| 2 | 2 | 1 |
| 3 | 1 | 0 |
| 3 | 2 | 0 |
| 4 | 1 | 0 |
| 4 | 2 | 0 |
Since there shouldn't be subject specific information in the design matrix, I am not quite sure how to proceed.
If I want to look at the interaction effect of time on group, while also correcting for age, would the matrix below be OK? ? In this example, the age is fixed at baseline. Should it be changing across time as well, such that the 2nd age for each is baseline age + scan interval years?
| Age |
| 41 |
| 41 |
| 30 |
| 30 |
| 51 |
| 51 |
| 40 |
| 40 |
Can you teach me how to create the contrast for this instance?
Thank you.
Best Regards,Shane
On Sunday, January 19, 2020, 10:24:02 PM GMT, Thomas Nichols <[log in to unmask]> wrote:
Hi Shane,
There are only some very specific mixed model analyses that PALM can do. Voxelwise, LME models are computationally intensive and can also have problems with convergence failures. I would suggest instead a related approach, using a sandwich estimator approach. In this approach, you only need to model the population effects (e.g. you don't have to specify if there are random intercepts or random slopes), ordinary least squares is used and then the repeated measures/longitudinal covariance is accounted for in the calculation of the standard error.
While I initially developed this as a SPM toolbox http://www.nisox.org/Software/SwE/ it is now available in FSL as "swe" https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Swe .
If you give swe a try, please let me and I'll be happy to help you get started.
-Tom
On Sat, Jan 18, 2020 at 10:20 PM Shane Schofield <[log in to unmask]> wrote:
Hi Anderson,
I have been using MATLAB for doing LME on ROI longitudinal data but now I would like to explore this using PALM. Apologies if I have missed the details but I can't see any LME option on the website. Specifically I want to test whether a variable is correlated with grey matter changes over time ( 2 time points), using random intercept for each subject while correcting for age and IQ. Any advice on how to implement this ?
Thank you.
Best Wishes,Shane
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
--
__________________________________________________________
Thomas Nichols, PhDProfessor of Neuroimaging StatisticsNuffield Department of Population Health | University of Oxford
Big Data Institute | Li Ka Shing Centre for Health Information and Discovery
Old Road Campus | Headington | Oxford | OX3 7LF | United Kingdom
T: +44 1865 743590 | E: [log in to unmask]
W: http://nisox.org | http://www.bdi.ox.ac.uk
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
