13.07.2015 Views

Contents

Contents

Contents

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CHAPTER 10: Quasi-Experimental Designs and Program Evaluation 329reason why we cannot conclude that groups are comparable based only onthe absence of a difference between the groups on the pretest. The pretest islikely to measure respondents on only one measure, or at best on a few measures.The mere fact that individuals do not differ on one measure does notmean they don’t differ on other measures that are relevant to their behaviorin this situation.Is there any reason to suspect a selection-maturation effect in the Langer andRodin study? That is, would it be reasonable to expect that residents on the treatmentfloor would change naturally at a faster rate than would patients on theno-treatment floor? Several kinds of evidence suggest that this would not be thecase. First, the procedure the nursing home used to assign residents to the twofloors was basically random, and the floors were assigned randomly to the treatmentand no-treatment conditions. Langer and Rodin also reported that theresidents of the two floors were, on the average, equivalent on measures such associoeconomic status and length of time at the nursing home. Finally, althoughit is not sufficient evidence in itself, residents on the two floors did not differ onthe pretest measures. Thus, the evidence strongly indicates that there was not athreat to the internal validity of the Langer and Rodin study due to the additiveeffects of selection and maturation.Selection-History Effect Another threat to internal validity that is not controlledin the nonequivalent control group design is the additive effect of selection andhistory. Cook and Campbell (1979) refer to this problem as local history effects.This problem arises when an event other than the treatment affects one groupand not the other. Local history, for example, could be a problem in the Langerand Rodin study if an event affecting the residents’ happiness and alertnessoccurred on one floor of the nursing home but not on the other. You can probablyimagine a number of possibilities. A change in nursing staff on one floor,for instance, might bring about either an increase or a decrease in residents’morale, depending on the nature of the change and any differences betweenthe behavior of a new nurse and that of the previous one. Problems of local historybecome more problematic the more the settings of the individuals in thetreatment and comparison groups differ. Langer and Rodin do not specificallyaddress the problem of local history effects.Selection-Instrumentation Effect A threat due to the combination of selectionand instrumentation occurs when changes in a measuring instrument are morelikely to be detected in one group than they are in another. Floor or ceilingeffects, for instance, could make it difficult to detect changes in behavior frompretest to posttest. If this is more of a problem in one group than in another, aselection-instrumentation effect is present. Shadish et al. (2002) point out thatthis threat to internal validity is more likely to be a problem the greater thenonequivalence of the groups and the closer the group scores are to the end ofthe scale. Because Langer and Rodin’s groups did not differ on the pretest, andbecause performance of the groups did not suggest floor or ceiling effects onthe measurement scales that were used, this threat to internal validity seemsimplausible in their study.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!