Power and Sample Size for Classification and Prediction

Sample size plays a key role in studies focused on classification and prediction. Whether you are puzzling about how many samples are needed for assessing diagnostic accuracy, or figuring out the smallest gap in receiver operating characteristic (ROC) curves you might be able to detect from existing data, it is important to be familiar with how sample size and power are related to the design of the study, effect size, and significance level. In this seminar we will discuss power and sample size considerations in the context of diagnostic testing, as well as classification and prediction problems, and will illustrate these relationships through examples. This seminar builds on concepts from Power and Sample Size 101. See Sample Size Computations for Medical Tests and Prediction https://ohsu-berd.shinyapps.io/pss_prediction/ To enter numbers and calculate sample sizes required for cases and controls.

Intermediate

Free