HSCI 330 Lecture Notes - Lecture 9: Confounding, Relative Risk, Japanese Americans
Document Summary
Learning objectives: distinguish between confounding and interaction. Identify interaction and effect modification in research results. Learning objective #1: distinguish between confounding and interaction (or effect modification). For our purposes, they are the same thing: two names for the same thing. To think about confounding, we calculate an adjusted effect estimate (like before) To think about confounding, we compare the crude and adjusted effect estimates. If they differ, then the c variable was confounding the crude effect estimate. If they are similar, then the c variable was not confounding the crude effect estimate. Here, there is a difference of about 16% between the crude and adjusted risk ratios ((1. 9-1. 6)/1. 9), so i would say smoking was confounding the crude lung cancer risk ratio comparing coffee drinkers to non-coffee drinkers. If there is effect modification, we usually don"t report an adjusted effect estimate. In general, if you have effect modification (or interaction) in your results, you should not report the adjusted effect estimate.