Date of Award

1-1-2011

Document Type

Dissertation

Degree Name

Ph.D.

Department

Quantitative Research Methods

First Advisor

Duan Zhang

Keywords

formative measurement, reflective measurement, social participation

Abstract

Social participation in late adulthood through activities such as volunteering with charities, playing sports, and joining clubs can increase life satisfaction, directly by providing enjoyable engagement and indirectly by increasing a person's sense of social connectedness. When reported levels of different types of activities are used to measure social participation, conventional measure development procedures based on classical test theory lead to a proliferation of small participation subscales that don't show good reliability, don't have theoretical power, and don't match researchers' conceptions of the dimensions of participation. Based on the poor performance of conventional approaches, some researchers have suggested that social participation should be modeled as an index composed of its indicators rather than as a scale in which indicators reflect an underlying latent factor. Typical approaches in psychosocial research rely on reflective-indicator models, which correspond to scale development, rather than incorporating composite variables with causal indicators. The latter approach, where manifest indicators are specified as causing the unobserved construct, is sometimes known as formative measurement, since the construct of interest is formed by its indicators. This study compared a scale model of social participation based on reflective measurement to an index model based on formative measurement.

Using a sample representative of community-dwelling U.S. adults over age 65 from the Health and Retirement Study's 2008 wave of data collection, two alternative measurement models of participation were constructed using sixteen items that recorded frequency of participation in different activities. Because patterns of participation differed for males and females, gender-specific models were developed. The scale models assigned participation items to subscales based on item intercorrelations. The index model assigned items to participation composites based on predictive associations with the outcomes of social connectedness and life satisfaction.

The index construction process led to a unidimensional representation of participation, composed of six of the original sixteen participation activity items. The initial attempts to build a scale model led to structures with many small factors and poor predictive validity. Based on the findings of unidimensionality for the index model, a single-factor scale model was explored for female respondents only. Results showed that both index and scale approaches have the potential to produce participation models that are parsimonious, well-fitting, and externally valid even though conventional scale development rules-of-thumb and current conceptions of the domains of participation lead the researcher to non-parsimonious, poorly-fitting solutions lacking predictive capability.

Participation measurement instrument developers often theorize the existence of three or more dimensions of participation. Whether they use conventional (reflective indicator) or more radical (formative indicator) models, they are advised, based on this study's results, to evaluate a single-dimensional structure among their candidate models.

Provenance

Recieved from ProQuest

Rights holder

Anne Zelenka

File size

253 p.

File format

application/pdf

Language

en

Discipline

Statistics, Gerontology, Psychology

Share

COinS