PSY 2311st Edition Lecture 21Outline of Last Lecture I. How Research Informs Clinical EthicsII. Traditional PsychotherapyIII. New Accountability for Treatment EffectivenessIV. Evidence Based Practice MovementV. Does Count as EvidenceVI. Why Therapists Should Care About “Bad Evidence” ReputationOutline of Current Lecture II. Effect SizeCurrent LectureEffect Size●Effect Size What Percent of Group 2 Beats Group 1’s MeanPercent of the Experimental Group That are HarmedPercent Helped0 50% 50% 0%.2 58% 42% 8%.5 69% 31% 19%.8 79% 21% 29%1 84% 16% 34%● 2 ways for results to be statistically significant○ almost always true-- really big mean difference + limited overlap○ small and unreliable -- small mean difference + much overlap really big N● Not obvious from p value○ percent of treated individuals who score better than the control group● Goal○ understand experimental results in terms of reliability effect○ challenge = error variance ( mean difference is big enough to be impressive despite overlap)● A standardized tool for weighing mean difference against variance○ effect size statistics = Cohen’s d○ d= M 1−M 2SD ○ SD for whole sample● Big t= small p● t= As meandifference gets bigger , t gets biggerAs SD gets bigger , t gets smaller / As N getsbigger , t gets bigger● BIg effect size = mean difference stands out, almost always true effect● Small effect size = error variance stands out, small and unreliable effect● The bigger the effect size, the more likely that an effect applies to most individuals● effect size-- measures reliability of the
View Full Document