Looking for a tool to assist with univariate exploratory analysis as an early step in a model building process.
Capabilities I’m looking for are to recode or transform the raw predictor variable, x, to produce a representation of x that is efficient for use in a model. Some examples are reducing levels of a categorical predictor to just the predictive ones, plus “other”, or determining if a smoother would be appropriate for a continuous predictor. It’s particularly important that the tool provide informative visualizations, test holdout sample performance, scale to a large number of predictors, handle different types of response variables, and produce a nice summary of the entire analysis.
Should I build it (in R), buy it, or borrow it from something open source that already exists? Looking for advice.