CHYS 2P52 Lecture Notes - Lecture 9: Internal Validity, Time Series, Quasi
15 Nov 2016
School
Department
Course
Professor
Document Summary
Major advantage: can do studies otherwise not doable: when participants cannot be randomly assigned, difficult and low frequency populations, when it is unethical to assign groups. Major disadvantage: (cid:272)an"t trust any findings a(cid:271)out group differen(cid:272)es due to multiple internal validity problems. Posttest only non equivalent control group design. Two groups are measured after one group receives the treatment. Can be improved by: randomizing prior to assignment, matching control variables, adding a pretest and using a time series design. Pretest posttest non equivalent control group design. Two groups are measured before and after one group receives the treatment. Differences in degree of change measured along with differences in posttest. Can be improved by: randomizing and matching prior to assignment. Using time and known/controlled events to sort out causality. One group is measured for time, then a treatment is introduced, then the measurement continues. If change occurs in synchrony with the event, then the event caused the change.
