Member-only story
How Biases Impact Causal Inference?
Exploring the Role of Confounding Bias, Selection Bias, and Measurement Bias in Causal Inference
Previous articles explored The Traps of Blindly Trusting Data: Simpson’s Paradox, An Intuitive Introduction to Causal Inference, and then Visualizing Causality Using Causal Graphs
This article delves into d-separation rules and explains different types of biases in causal inference: confounding bias, selection bias, and measurement bias, with examples.
Recap of Causal inference, Causal Graphs
Causal inference is the process of understanding cause-and-effect relationships. It focuses on inferring the causal effect of an intervention, treatment, or exposure on a particular outcome.
Causal graphs are Directed Acyclic Graphs (DAGs) used to model relationships between variables in causal inference. They depict expert knowledge and assumptions about how variables interact, aiding in understanding observations and drawing accurate conclusions about cause and effect.
There are three structural reasons for the two variables to be associated:
- Association: one variable causes the other. Association refers to a statistical relationship between two linked variables; when one changes, the other often does too.
Rain (A) causes wet ground (Y)
