Related Experiment Videos
Use of psychometric techniques in the analysis of epidemiologic data
Samuel F Posner1, Leavonne Pulley, Lynn Artz
1Division of Reproductive Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, GA 30341, USA. shp5@cdc.gov
Purpose:
This article demonstrates techniques for developing reliable multi-item scales for analysis of complex public health data.
Methods:
Information from a questionnaire designed to evaluate the acceptability and efficacy of the female condom as a method for STD/HIV prevention was summarized using psychometric analysis. 1159 high-risk women attending STD clinics participated in this study. Questionnaire items were designed to measure nine domains of predictors of condom use.
Results:
Principal components analysis was employed to reduce the number of potential predictors. Reliability of the multiple-item scales was assessed using Cronbach's alpha. Pearson's correlation coefficients were calculated to evaluate collinearity among multi-item scales. Approximately half (51%) of the questionnaire items that were analyzed were retained in the final scales. Data reduction procedures identified several multi-item scales with acceptable reliability (Cronbach's alpha >0.70). The correlation coefficients between scales was never >.5, suggesting that there was little collinearity among the scales.
Conclusions:
When focused on multiple partially interdependent determinants of an outcome, data reduction decreases the number of independent variables to be evaluated, ensures they have adequate reliability, maximizes strength of their association with outcomes, and reduces collinearity among predictors.