So with the EV1 which is in my case time measure in relation to the baseline (in days), this EV1 looks like :

 EV1

Time
0
136
0
28
0
427
0
71
0
168
0
399
0
63

With the contrasts [1 0 0 0 ...] and [-1 0 0 0 ...], how could I determine when Time1 > Time2 or Time2 > Time1 ?

 

Another possibility is to consider the sandwich estimator, available in an SPM toolbox by Bryan Guillaume and Tom Nichols, which bypasses some issues related to compound symmetry and missing data. You may want to read the documentation and see if it applies to your case.

Ok, thanks I will look at this sandwich estimator.

 

All the best,

Anderson

Best regards,

Matthieu
 

On 30 September 2015 at 08:47, Matthieu Vanhoutte <[log in to unmask]> wrote:

Hi Anderson,

Indeed I reduced my patients list because of quality of input images, and now I have 7 patients in one group with 2 up to 4 timepoints non regularly spaced in time :

- patient1: 2 timepoints
- patient2: 3 timepoints
- patient3: 2 timepoints
- patient4: 4 timepoints
- patient5:4 timepoints
- patient6: 3 timepoint
- patient7: 3 timepoints

Could you explain me how to use longitudinal TBSS with these data (define inputs, design matrix and contrasts) ?

Best regards,

Matthieu

2015-09-29 11:41 GMT+02:00 Matthieu Vanhoutte <[log in to unmask]>:

Hi Anderson,

I would like to investigate for example FA reduction with time. The number of timepoints is varied because of the clinical nature of the study: some patients came one time whereas others came up to seven times.

Moreover, time between two timepoints aren't regular between patients. I have 11 patients with the following timepoints:

- patient1: 3 timepoints

- patient2: 3 timepoints

- patient3: 2 timepoints

- patient4: 3 timepoints

- patient5: 7 timepoints

- patient6: 1 timepoint

- patient7: 3 timepoints

- patient8: 2 timepoints

- patient9: 7 timepoints

- patient10: 3 timepoints

- patient11: 3 timepoints

Do you think it is possible to make longitudinal TBSS on these irregular timepoints subjects and in this case how define correctly input and model for statistical analysis ?

Thanks in advance for helping !

Best regards,

Matthieu

2015-09-29 11:02 GMT+02:00 Anderson M. Winkler <[log in to unmask]>:

Hi Matthieu,

The "evolution" is too much a broad term. What exactly do you want to investigate? Could you explain the why the number of timepoints is varied? Is the missingness only towards the end, or are there gaps in the middle? Please, give as much detail as possible.

All the best,

Anderson

On 29 September 2015 at 09:49, Matthieu Vanhoutte <[log in to unmask]> wrote:

Dear FSL's experts,

I have one group and variable number of timepoints per subject (from 1 up to 7) and would like to assess the evolution according time-points of DTI parameters (FA, MD, ...).

Concerning the statistical analysis, how should I proceed in terms of inputs and models of the longitudinal variable number of timepoints ?

Many thanks in advance for helping !!

Best regards,

Matthieu