BookmarkSubscribe
Choose Language Hide Translation Bar
neurodoc
Community Trekker

How to set up an ANOVA for ROI analysis

Apologies in advance to seasoned JMP users; this is a newbie question. 

I have a dataset of anatomically-derived brain ROIs measuring fractional anisotropy (FA). There are a total of 20 ROIs, some of which are 2nd order (derived from the others). My main factor is group (normal vs dementia). For the FA analysis, I would also like to control for size differences between the SAME ROI in the 2 groups (I have ROI size, in mm for each one). Also age and sex. How do I set up the ANOVA in JMP13? Data structure is as follows:

 

Subject 1: group, ROI1_FA, ROI2_FA, ROIn_FA, ROI1_size, ROI2_size, ROIn_size, age, sex

Subject 2: group, ROI1_FA, ROI2_FA, ROIn_FA, ROI1_size, ROI2_size, ROIn_size, age, sex

Subject x: group, ROI1_FA, ROI2_FA, ROIn_FA, ROI1_size, ROI2_size, ROIn_size, age, sex

 

0 Kudos
6 REPLIES 6

Re: How to set up an ANOVA for ROI analysis

Are the ROI measurements for each subject considered independent responses?

Learn it once, use it forever!
0 Kudos
neurodoc
Community Trekker

Re: How to set up an ANOVA for ROI analysis

No they're not. These ROI measures will be highly correlated, as they are different-but-adjacent brain regions from the same individual. 

0 Kudos

Re: How to set up an ANOVA for ROI analysis

Thanks for your reply. This note should be helpful to you in setting up your analysis. There are three different approaches for you to consider.

As for the size, I don't know enough about your field to suggest anything but the most obvious normalization approaches, such as ratio (ROI / size). You might be able, though, to use it as a covariate if you use the 'split-plot' approach described in the note.

Learn it once, use it forever!
0 Kudos
neurodoc
Community Trekker

Re: How to set up an ANOVA for ROI analysis

Thank you, that article was definitely helpful. I am certain that what I was looking for was the Mixed Model "personality," and what I was missing was knowledge of the "stack" command under the Tables menu, which allowed me to restructure my data to fit the method. I am now faced with another, higher-level question, which is the choice of repeated covariance structure for my data. I suspect I need one of the 4 "spatial" options, but this is just a hunch, and I certainly don't know which of the 4 to select, nor which "Type" is best for these data, nor what is meant by the required "Repeated" input. Is there a page that discusses these options?

All in all, what would be most helpful would be to find a discussion somewhere of how to use JMP with MRI (or similar imaging) data. Is there such a page, or group available?

0 Kudos
neurodoc
Community Trekker

Re: How to set up an ANOVA for ROI analysis

bump. 

0 Kudos
jiancao
Staff

Re: How to set up an ANOVA for ROI analysis

You do not seem to have the spatial information to fit a spatial model using Mixed Model. Here is an example of fitting such model. https://www.jmp.com/support/help/13-2/Spatial_Example_3a_Uniformity_Trial.shtml#1264182

 

However, with the data you've collected shown above and stacked where there are Subject column identifying subjects and Repeated column identifying the sequence of ROI_FA, you can run the analysis of repeated measures using Mixed Model. ROI_size as well as age and sex enter the model as fixed effects. To start, try some of the simpler covariance structures https://www.jmp.com/support/help/13-2/Fit_Model_Launch_Window_2.shtml