PSY 370 Lecture Notes - Lecture 17: Regression Analysis, Job Analysis, Psy
Document Summary
Using more than one predictor to predict the same outcome. If the predictors provide new information (are not highly correlated): Adding the second predictor doesn"t improve prediction much. Your items are relevant to and cover all of your construct. Degree to which items are representative of the construct the test is intended to measure. No quantitative analysis of item responses can do this for you. A well-documented testing universe & test blueprint can be evidence for appropriate content. Based on a thorough review of relevant theory and/or research. In io, this would be a job analysis! Clearly show the links between specifications in your test blueprint and the items you wrote. Sounds easy, but requires a lot of work. People knowledgeable about the content of the test. In addition to, not instead of, paying attention to content validity during development. People who are knowledgeable about the content of the test. Content validity ratio (we did this in lab!)