PSY 3711 1st Edition Lecture 17Outline of Last Lecture I. Disparate treatmentII. Disparate impactIII. Us legal backdropIV. Civil rights act of 1991V. Americans with disabilities act of 1990VI. Characteristics of medical screenVII. Equal employment opportunity commission (EEOC)VIII. Testing and the civil rights actIX. Clinical synthesisOutline of Current LectureI. Humans behind the machineII. Meta analytics weightsIII. Expert WeightsIV. Bootstrapped weightsV. Which is best?Current LectureI. Why does mechanical data combination perform so wella. We can mechanically combine predictors in a number of different waysb. Equal unit weightsc. Differential weightsi. Expert judgment based weightsii. Bootstrapped weightsiii. Criterion weightingII. Meta analytic weightsa. Advantage of using more stable estimatesb. Can be used in two waysi. Use the predictive validityii. Use both the predictive validity and the info about its interrelations with other predictorsIII. Expert weightsa. For expert judgment based weights we do a study of experts and ask them how they weight the infob. We simply ask them how they do itThese notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.c. Only as good as the experts askedIV. Bootstrapped weightsa. Try to model what experts do and then use hteir weighting scheme consistentlyb. One way to do this is to gather the judgments of a large nubmero f experts and then “predict” their judgments based on the data they had availablec. Much like optimal weights, a multiple regression is usedV. What predictors are besta. When predictors are of similar strength, and there is not a lot of overlap, unit weights work just fineb. Difference between equal unit weights and differential weights is typically smallc. If there are large difference in predictive strength of your measure, then differential weights can yield some improvementd. But even if you only use unit weights, you are already way
View Full Document