Individualized conformal prediction: using synthetic data as relevant controls

Abstract

The problem of individualized prediction can be addressed using variants of conformal prediction, obtaining the intervals to which the actual values of the variables of interest belong. We present a method based on detecting the observations that may be relevant to a given question and then, based on them, generating simulated controls to yield intervals for the predicted values. This method is shown to be adaptive and able to detect the presence of latent relevant variables.