Evaluation of factor structure of a nicotine coping study via generalized latent variable modeling and cross validation techniques

Chris M Manuel, The University of Texas School of Public Health


A common strategy in smoking cessation therapy is to teach skills for coping with the craving to smoke. Often, it is of interest to determine what kind of coping strategies participants have employed before, during, and after the intervention. A questionnaire, the Coping Behaviors Scale, was developed to measure a smoker's self-identified adaptive and maladaptive coping strategies. We sought to produce a statistically validated questionnaire by conducting formal psychometric analysis, chiefly factor analysis, on a baseline sample of ordered categorical data, with a size of n=877 patients pooled from three smoking cessation clinical studies. Model formulation for our data was based on the General Structural Equation Modeling utilized in the MPlus software. Model parameters were estimated using the Weighted Least Squares (WLS) technique. A cross validation like technique, which split the sample into three datasets with equal numbers of each study, was employed to ensure the replicability of the factor structure. Scree plot results indicated 3 and 4 factor solutions. Initial results showed that the cross validation technique did not provide the best fit statistics; however, issues with one of the variables made a second round of analysis necessary. As with the first analysis, scree plot results indicated 3 and 4 factor solutions as most appropriate. Among all the models, we found that the 3 factor solution using the cross validation like technique provided the best fit and interpretation. Based on the content of the items that comprised each factor we named two of the factors Proactivness, Resolve, and Resignation. Finally, we explored potential relationships between the factors and various sample characteristics such as gender, ethnicity, age, type education and employment status. Because analysis of the residuals exhibited high skewness, we employed a Negative Binomial Regression to assess the possible associations. We found potential relationships with African American ethnicity, age, education, and WSWS Craving Score. These analyses resulted in a psychometrically valid measure of coping behaviors that can be used in smoking research and will provide insight for identifying individuals who may require more rigorous cessation therapy than other participants.

Subject Area

Biostatistics|Public health|Quantitative psychology

Recommended Citation

Manuel, Chris M, "Evaluation of factor structure of a nicotine coping study via generalized latent variable modeling and cross validation techniques" (2014). Texas Medical Center Dissertations (via ProQuest). AAI1568968.