"Overview of methods for handling missing values".

Abstract

The problem of missing values exists since the earliest attempts of exploiting data as a source of knowledge as it lies intrinsically in the process of collecting, recording, and preparing the data itself. It is all the more unavoidable as vast amounts of data are currently collected from different sources: "One of the ironies of Big Data is that missing data plays an increasingly important role". There is a vast literature on this topic, and a recent survey even identified more than 150 different implementations.

In this presentation, I will share my experience on the topic. I will start by the inferential framework and then show how missing values create additional challenges to the task of supervised learning, as traditional machine learning algorithms can not handle incomplete data. Finally, I will illustrate the impact of the methods developed in the causal inference field to estimate treatment effects from clinical data.