Let’s learn about Goodman-Kruskal’s Lambda (GKL). This association measure is a Proportional Reduction in Error (PRE) measure. A PRE measure determines the reduction in the probability of error of one variable given another. GKL is an asymmetric measure; for two categorical variables, A and B, Lambda of A given B is not equal to Lambda of B given A. In plain words, for two variables, A and B, GKL can help to determine if A is a better predictor of B or vice-versa. This asymmetry property has implications for causal researchers, as causality is looked upon as having asymmetry between the cause and effect. We use py-bbn to simulate data and py-pair to compute GKL in this video. Both of these libraries are open-source Python modules available from PyPi.

oneoffcoder

0

Tags :