Let's say I have a small data set (~50 observations) and 50 or so explanatory variables, and I want to develop a linear model for the sake of prediction. First, I want to construct a correlation matrix or a variable selection technique to select a smaller number of predictors to include within the model. However, I'm wondering what would be the appropriate action when attempting to train the model and checking whether it has predictive power? train-test data split? cross validation? what would be the best could of action with a small dataset and large number of explanatory variables?
Tagged w/ R because I'll be using R to conduct analysis.
