Dear community,
I am aiming on identifying patterns from a multi-omics dataset (10 000s of metric variables per sample) that robustly associate with certain clinical outcomes (binary). I am familiar with techniques like PCA etc. that allow for identifying patterns within the dataset but am specifically looking for an approach that allows for sample size calculation and a-priori definition of clear hypotheses that can be tested. Our sample is rare, thus we have approx. 400 multi-omics samples (10 000 variables each) to correlate to 250 clinical observations (patients) with 4 recorded outcomes in binary/ordinal scale.
I would really appreciate any ideas that would allow me to perform a robust "non-exploratory" statistical analysis for this study.
Yours sincerely,
Michael
If the analysis is "non-exploratory", as you say, then you have a particular statistical model. What hinders you to apply your data to it?
You should look into regularized structural equation modeling.
Article Regularized Structural Equation Modeling
It is made specifically for the kind of problem you are proposing.
Robert Moulder thank you very much for your suggestion - I will look into it.
In the meanwhile our statistician proposed opting for this approach instead:
https://academic.oup.com/bioinformatics/article/21/19/3771/210591?login=true
I´d love to hear your opinion on it!
- Michael
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity