Add spline features for continuous variables

[schuemie]

Some of the non-binary variables such as age and Charlson index are currently provided as their verbatim value, which means models such as logistic regression (propensity scores) will model them as linear. However, a linear assumption is almost never realistic.

FeatureExtraction could also offer these same variables as splines, by already computing the spline design matrix. An example where I've done this before is here in the SelfControlledCaseSeries package. The hard part would be the administration of the covariate IDs for the design matrix variables.

[jreps]

This would be great! PLP has spline feature engineering code, but if FE can do it that would be excellent!