Hi,
I'm trying to investigate if there are any relationships between my variables (risk factors) and anxiety score for participants of my study. Unfortunately I only have 24 participants/observations. I have 8 risk factors/variables but they are categorical so I have dummy coded them but this means I end up with 24 variables and so cannot run a MLR. Is there any way around this, or any other way that you can run a correlation or MLR with the same number of variables as observations? TIA
You can run a regression with the number of observations equal to the number of variables, but it is guaranteed to have a perfect fit without regard to whether there is any relationship. So it really isn't useful. (I assume that some of the risk factors have more than two categories.)
You might like to look at the lasso, which can combine feature selection with fitting. It can be a steep learning curve, but start with Lasso (statistics) - Wikipedia, and then look at the section titled "Elastic net". The R package "glmnet" (CRAN: Package glmnet) is the standard implementation. It might work for your problem.