8.3 Some Remarks and Questions

For LASSO, we are using the basis of polynomial splines of degree 2 with interior knots at 25th, 75th, and 50th percentiles of each feature. We can see that although the effects are picked up, its slightly late and are lower compared to the true effect. A basis for linear splines performs well in this case.

The causal forest framework on the other hand performs better.

Rubin, Donald B. 1974. “Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies.” Journal of Educational Psychology 66 (5): 688.

Splawa-Neyman, Jerzy, Dorota M Dabrowska, and Terrence P Speed. 1923 [1990]. “On the Application of Probability Theory to Agricultural Experiments. Essay on Principles. Section 9.” Statistical Science, 1923 [1990], 465–72.