Studies of behavior in the workplace have typically relied on quasi-experimental designs, primarily due to the impracticality of implementing a randomized experiment in vivo in a workplace. Instead, nonequivalent comparison group designs are typically used–e.g., provide an intervention for the employees of Plant A but not those of Plant B–and analysis of covariance (ANCOVA) models are used to “adjust” for baseline differences between the groups. Recent advances in the methodological literature, however, have shown clearly the limitations of ANCOVA in making the groups “equivalent.” As a result, the confidence that can be placed in inferences about the efficacy of the intervention studied is limited. Using data from sites participating in the CSAP Workplace Managed Care Program–a multisite collaborative created to study substance abuse prevention in workplaces with managed health care–we show specific examples of the limitations of ANCOVA, and demonstrate the value of the propensity score approach to the analysis of nonexperimental data with respect to causal inferences about intervention efficacy. Session participants will be able to list the conditions under which ANCOVA models are likely to be ineffective, to recognize the advantages of the propensity score approach, and conceptualize propensity models.
Learning Objectives: Session participants will be able to list the conditions under which ANCOVA models are likely to be ineffective, to recognize the advantages of the propensity score approach, and conceptualize propensity models.
Keywords: Worksite, Substance Abuse
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
Disclosure not received
Relationship: Not Received.