You’ve all heard about Machine Learning. Some say it’s amaying, I say it’s meh. I would rather double it and give it to the next person which, happend to be you (see the pun I just did ?). You guessed it today we will talk about Double Machine Learning.

NB: For now we will focus on how Double ML is used to compute ATE. CATE may come later.

Introduction

Before diving into this subject we need to talk about causation and causal inference.

We will start with a very basic example:

Say we have two populations $P_D$ and $P_C$ where people in $P_D$ were given a drug to increase their heights and people in $P_C$ were not. A few weeks lates, we measure the average height ($H_D$ resp. $H_C$) of each population and find out that $H_D$ > $H_C$. Wait what ? Does this mean that the drug works ? Let’s rewind a bit.

We travel back in time a few weeks before people in $P_D$ were given the drug, measure their initial height $H_D’$, and compare it to $H_C$. We make a shocking discovery: $H_D’$ > $H_C$.

Our initial thought was equivalent to the following computation:

If we expand this calcul we get