Hi experts,
I am new NBSer. I have two groups, one is ctr, another is intervention. After  2 years followup, I would like to check 1) is there any difference in network at baseline and followup between two groups, 2)  if the change in network during the 2 years differed between the two groups, and use age as a confounder. How to design the matrix?
Thanks in advance,
Best regards,
Yawu

Hi Yawu,

examples of this kind of design can be found in previous posts on this forum.

In brief, your design matrix should comprise a column for the main effect of group (1=patient; 0=control), another column for the main effect of time (0=first time point; 1=second time point), another column for age.

Finally, if the follow-up period involves the same subjects as baseline, you should consider a repeated measures design and add a separate column for each individual, where each column will contain two 1's indicating the individual, with 0's elsewhere.

Andrew



Originally posted by Yawu Liu:

Dear Andrew,

Thanks for quick reply. I am digesting, and will consult again if any new questions emerge. 

Yawu

Hi Andrew,

I have a doubt regarding longitudinal analysis in NBS toolbox. The data I am working on is NICAP data, where I am looking at three time points. I have  extracted three connectivity matrix (using MMP atlas (360 regions)) for each of the wave separately and we have missing data over time. I am trying to understand the significant connections across these nodes over time. I wanted to know whether I have to convert the data into Long format to compute longitudinal analysis. I was wondering  considering repeated measures for each participant for 360 regions is time consuming.Is there  any  other way to compute this? Also, what can I do for missing data?

Thanks and Regards,
Shania

Hi Shania, 

The numeric format of the data (i.e. long vs short) is not related to the longitudinal nature of an analysis. There is not relation between Long format and longitudinal analyses. 

360 nodes should be feasible and not too time consuming, unless the sample size is particularly large. 

If the number of subjects with missing data is small, you may want to exclude those subjects. Otherwise, a mixed effects model can be used to account for missing data, although this is not available in the current version of NBS. 

You can run a 3x1 ANOVA and an F-test to assess for differences between the three time points. Check out the Help section for some examples. The FSL FEAT webpage provides some useful design matrix examples as well. 

Good luck!

Andrew



Originally posted by Shania Soman:

Thank you for the reply! 

I also have doubt regarding NBS analysis using nbr_lme function in R. Sorry, that it is a very silly question. I am very new to this analysis in R. It will be very helpful if you can help me in this.

I have 3 functional connectivity matrix of size 360x360x146 , 360x360x136, 360x360x96 for each three wave respectively. Could you please let me know how will I input the matrix/ data frame to the argument "net" to do nbr_lme analysis. How can I input 3D volume in long format?

Many Thanks,
Shania

Hi Shania, 

For the R version of NBS, please contact Zeus Gracia-Tabuenca for assistance:

zgtabuenca at comunidad.unam.mx


https://cran.r-project.org/web/packages/...

Originally posted by Shania Soman: