0
$\begingroup$

I have two datasets from patients with different diseases, which contain multiple measures of regional brain function and performance on different memory tasks. The memory tasks used in each dataset were not exactly the same, but are meant to assess the same dimensions of memory. My goal is firstly to see which memory task scores correlate with function of different regions. Secondly, I want to compare across different disease to see how similar they are to each other.

For my first goal I am thinking to do a partial least squares correlation, within both datasets (seperately), giving me pairs of correlation values for tasks and brain regions. For the second I would like to do something like cluster these correlation values, which would tell me, for example, disease A has working memory problems and dysfunction of X brain region, whilst disease B has long term memory problems and dysfunction of Y brain region.

The diseases in dataset 1 are very specific, there are 3 and they should match on to some distinct subtypes, so I could cluster those. The diseases in dataset 2, I don't know which tasks/brain regions they will have problems with.

My questions are:

  1. Does the PLSC make sense?
  2. Does clustering based on the PLSC values make sense?
  3. How can I "compare" the datasets? Could I cluster them separately and compute a distance metric, or pool them and see which cluster the patients from dataset 2 end up in?

Thank you for reading! Any help greatly appreciated!

Improve this question
$\endgroup$
2
  • $\begingroup$ I'm struggling to understand the specific hypothesis. You have multiple datasets, each dataset has a different memory task, and brain imaging, and you'd like to know if the memory task from one dataset can predict the disease status as if the data were pooled with a uniform memory task performed on the combined sample? $\endgroup$
    – AdamO
    Commented Sep 8, 2023 at 17:24
  • $\begingroup$ Thanks for your comment, sorry I just saw it. The hypotheses would be, 1) performance on different memory tasks correlates with brain function of different regions, and 2) these latent associations can be used to cluster different diseases. I hope that's clearer. I'm just not sure how to work with the two datasets $\endgroup$
    – firefly
    Commented Sep 11, 2023 at 14:33

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.