PHI 31 Lecture 24: phi31 - lecture 24 and 25
Document Summary
Red: experimenter randomly assigns subjects to treatment and control groups that control the entire residual state. Pro: can"t meet random sampling assumptions as well as red. Only a few aspects of residual state can be controlled or matched. Nature or self-selection assigns subjects to treatment or control groups, which could be correlated or causally interact with residual state in an unknown way. Ret: both cause and effect have appeared before the study begins. Study gathers cases with the effect, constructs controls to match cases on residual states (to stimulate random sampling), then searches background/past for signs the cause is present or absent. Have little chance of true random sampling. Eliminating bias partially addresses lack of random sampling. Random sampling is ret is at best somewhat well met . Look for efforts to match subjects and eliminate bias in report of a ret. Review of causal studies and decision concepts and models of decision-