Using multilevel modeling in caregiving research.

Recently, caregiving research has moved towards a greater emphasis on the interactive and dynamic processes of the care situation. As a result, many of the questions are multilevel in nature, involving more than one unit of analysis and placing greater focus on the context of the care dyad. Nevertheless, most caregiving studies use methods that fail to address the hierarchical nature of the care dyad and consequently fail to illuminate variation both within and between dyad processes. The focus of this paper is to promote a dyad approach to the analysis of caregiving data and to highlight the advantages and appropriateness of multilevel modeling when dealing with clustered or hierarchical data. In particular, we discuss how multilevel modeling can be adapted to study change in the care dyad over time, improve the way we evaluate interventions, and enhance our understanding of dyad congruence.

Perceived exposure to peer and adult drinking as predictors of growth in positive alcohol expectancies during adolescence.

This study investigated the effects of change in exposure to peer and adult drinking on changes in positive alcohol expectancies during adolescence. Covariance and mean structure analysis were used to model change in the predictors and in alcohol expectancies in a sample of southern California schoolchildren followed from Grades 5 to 10 (N = 3,580). The sample was gender balanced and was predominantly White (51%) and Hispanic (28%). Exposure to peer drinking and rate of change in exposure to peer and adult drinking were found to predict the rate of change in alcohol expectancies. These effects were particularly strong between Grades 5 and 7, suggesting the need for early intervention aimed to prevent the increase in positive alcohol expectancies.

A Cross-Domain Model for Growth in Adolescent Alcohol Expectancies.

Methodologists have recently shown how the methods of individual growth modeling and covariance structure analysis can be integrated, bringing the flexibility of the latter to bear on the investigation of inter-individual differences in change. The individual growth-modeling framework uses a pair of hierarchical statistical models to represent: (a) individual status as a function of time, and (b) inter-individual differences in true change. Under the covariance structure approach, these level- I and level-2 models can be reformatted as the "measurement" and "structural" components of the general LISREL model with mean structures. Consequently, a covariance structure analysis of longitudinal panel data can provide maximum-likelihood estimates for all level-2 parameters. In this article, using longitudinal data drawn from a school-based alcohol prevention trial, we demonstrate how the new approach can be used to investigate the inter-relationships among simultaneous individual changes in two domains - positive arid negative alcohol expectancies - over the course of early to mid-adolescence, for both boys and girls. We represent individual change over time in positive expectancies with a piecewise growth model, and in negative expectancies with a straight-line growth model. Then, we use multi-sample covariance structure analysis to ask whether individual changes in positive and negative expectancies are related to each other and whether the pattern of inter-relationships differs by gender. Our approach can easily be generalized to more than two domains and has a variety of other advantages that we document in the discussion.

Sticks and stones may break my bones ... reasoning about illness causality and body functioning in children who have a chronic illness.

Children's concepts about illness causality and bodily functioning change in a predictable way with advancing age. Differences in the understanding of these concepts in healthy children vs children with a chronic illness have not been clearly delineated. This study included 49 children with a seizure disorder, 47 children with an orthopaedic condition, and 96 healthy children, all with normal intelligence and ranging in age from 5 to 16 years. It demonstrates systematic differences in children's general reasoning skills and in their understanding of concepts about illness causality and bodily functioning, as a function of their age and experience of illness. At all ages, children who had a condition with orthopaedic involvement reported less sophisticated general reasoning and concepts about illness than did healthy children; children with a seizure disorder reported similar general reasoning skills to those of healthy children, but considerably less sophisticated concepts about illness. Children's concepts about body functioning did not differ as a function of the presence of a chronic illness. When their different levels of general cognitive reasoning were statistically controlled, children with a chronic illness had somewhat more sophisticated concepts about bodily functioning than did healthy children. Differences in conceptual development among children with different types of illnesses lead to interesting speculations with regard to the effects of particular illness characteristics on children's cognitive development.