Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Usree Kirtania, MS and Cynthia Davis, MS. Institute for Community Health Promotion, Brown University, 1 Hoppin Street, Coro 4 West, Providence, RI 02903, (401) 793-8342, ukirtania@yahoo.com

The importance of a well-constructed scale from a questionnaire is essential for testing its reliability and validity. To achieve this, clear understanding of underlying dimensions of questionnaire variables is necessary. The aim of this presentation is to identify conceptually interpretable dimensions of a questionnaire using a variable cluster analysis approach (SAS PROC VARCLUS). We illustrate this approach using a Food Habit Questionnaire (STFHQ) to capture dimensions of fat related eating behaviors and compare this with factor analysis findings as well as existing factors from pilot study. STFHQ is a 94-items food habit questionnaire that was implemented on a weight control intervention program (SisterTalk) for African American women. To achieve non-overlapping groups of variables (cluster) that are relatively highly correlated, VARCLUS uses iterative splitting and factor analytic method. Associated with each cluster is a linear combination of the variables in the cluster, the first principal component. Clusters are chosen to maximize the variation accounted for by the first principal component of each cluster. The findings about which items are most closely related are almost similar in VARCLUS and factor analysis methods compared to existing factors from pilot study. This method developed simple, distinct and easy to interpret dimensions while factor analysis method showed some overlapping factors. The observed clusters of variables empirically supported most of the conceptual dimensions of the fat related behaviors. Variable cluster analysis seems a useful approach to capture the different dimensions of a questionnaire than factor analysis method in terms of simplicity and interpretability.

Learning Objectives:

At the conclusion of this session participant will be able to discuss
• The methods and uses of variable cluster analysis.
• The differences between variable cluster analysis and traditional factor analysis method.
• The advantages of variable cluster analysis when creating a scale from a questionnaire.

Presenting author's disclosure statement:

Not Answered

Handout (.ppt format, 633.0 kb)