Overview

While correlation is useful for prediction sometimes the question that’s being asked is if a variable or factor actually causes a result. In order to gain experience in casual analysis or casual inference I set out to build a layered model that would look through a data set and find casual relationships. While I could build the model, finding data I can perform analysis on has been a struggle. This is due to the fact that causal inference has a lot of assumptions that must be accounted for in order to make the jump from correlation to causation. So until I find data that I can make proper assumptions on, I will only leave the outline for my model below. ( code is also linked in below)

Link to Code

Blueprint

General overview

General Propensity score model
Outcome model
Causal inference model
*Each model is then used to build the next

Models used for General Propensity score

Linear Regression
Support vector regression
- Hyper parameter optimized
Random Forest
- Hyper parameter optimized

Models used for outcome prediction

Logistic Regression
Boosted logistic regression
Random forest

Models for Casual

G-computation
Double machine learning (not-implemented)

Link to Code