Which statistic test is it correct to use to detect common variable within a group of sample?

Question

[AQ7]

Good morning,

I'm currently on cancer research and i have several variables for a group of cancer patients. I would like to know which test do I have to apply to detect which of these variables are commons among all patients of my cancer group, so to detect multiple variables that define that group? Do I have to perform a Randome Forest in this case? thank a lot for your help

[boczniak767]

The CrossValidated forum is more appropriate for this question (https://stats.stackexchange.com/) and you'll get fasterresponse here

[AQ7]

ah ok I'll use that forum thanks

[Jean-Karim Heriche]

Your question is not entirely clear to me. What do you mean by detecting common variables ? What does your data look like ?

[AQ7]

well imagine that i have a excel where each column correspond to a clinical value and in rows you have cancer patients and control patients. So i would like to know if one or more than one of these variable is like a "biomarkers" for the cancer patients.

Answer 1

[Jean-Karim Heriche]

To identify variables (aka features) associated with one class, look for feature selection/extraction methods in machine learning. I would start with PCA and/or LDA. You could also try a generalized linear model and check the contributions of the variables.

[AQ7]

thanks a lot for your sugestion. could you please suggest me one of these linear model you are talking about thanks

[Jean-Karim Heriche]

Have look at logistic regression e.g. with the glm() function in R.