Psychology 2300 Chapter 10 More on Experiments Pre Post Tests One group of participants Measures dependent variable before and after treatment May lack control of other variables that can produce confounds Threats to Internal Validity History event confounded with treatment o Example treatment for managing money during a recession Maturation growth natural changes in participants society over time o Longitudinal and successive independent samples Testing threat taking pretest affects post test Regression extreme scores on pretest are less extreme on post test o Example foreign language placement test Attrition people drop out of the study same as subject loss Instrumentation changes in the instruments over time o Example DSM criteria in 5th edition Contamination the treatment group members talk to the control group members so even the control group can practice on themselves diffusion of treatment Demand Characteristics when participants are aware of what the researcher expects to find which may cause them to alter their behavior to conform to expectations o Can be prevented by using a double blind study Observer bias occurs when the observer knows the goals of the study and allows it to influence their observations Placebo effect a positive effect in a patient following a treatment which arises from the patient s expectations rather than the treatment itself Definition of Quasi Experiment Includes an intervention or treatment Includes measures of the dependent variable taken after and in some cases before the treatment Lacks random assignment to conditions but may include a comparison group May lack control of other variables that can produce confounds The Logic of Quasi Experiments Don t let the perfect drive out the good o It s better to do the study and recognize the flaws than to not do the study Need group to serve as comparison group Quasi Experimental Design Categorization O measures of dependent variable before pretest and or after post test the treatment X treatment applied by researcher another agent or naturally occurring When O s and X s are on 2 or more lines and separated by dashes multiple groups with no o A good example would be people attending a Weight Watchers support group weigh them random assignment Read left to right One Group Pretest Post Test Design O1 X O2 before and after a 10 week support group Nonequivalent Control Group Designs Nonequivalent control group WITHOUT a pretest o X O O o O1 X O2 O1 O2 Nonequivalent control group WITH pretest Interrupted Time Series Designs any design with multiple measures of the dependent variable taken over time before and after treatment Basic design o O1O2O3O4O5O6O7 X O8O9O10O11O12O13O14 Often uses archival measures due to number of data points needed over time Archival data data that s already been collected Look for abrupt changes aka discontinuities from immediately before to immediately after the treatment Threats to internal validity o History is a major threat did anything else happen around the same time o A series of pre and post measures controls the threats of maturation testing threat and regression Interrupted Time Series With Nonequivalent Control Group Design a combination of the two major designs Basic design o O1O2O3O4O5O6O7 X O8O9O10O11O12O13O14 O1O2O3O4O5O6O7 O8O9O10O11O12O13O14 responsible for treatment effects Helps to deal with threat of history in that an event that affects both groups cannot be