|
Abstract:
|
This talk is based on three consulting projects that I did when I was working with investigators in a medical school. All the three projects needed variable selection, but because they had different goals, I applied different variable selection procedures. Specifically, I used Variable Importance for the goal of prediction, AIC for the goal of explanation, and Propensity Score for the goal of Verification. Although modern variable selection procedures such as LASSO are dominating the literature, I found the above three variable selection procedures are useful for communicating with investigators in medical schools.
|
