Multivariate statistics for family psychology research.

James H. Bray,Scott E. Maxwell,David Cole

DOI: https://doi.org/10.1037/0893-3200.9.2.144

1995-01-01

Journal of Family Psychology

Abstract:The application of selected multivariate statistics is illustrated for use in family psychology research. The use of multivariate analysis of variance (MANOVA) and discriminant analysis in factorial designs and profile analysis is discussed. Profile analysis provides a method for dealing with unit of analysis issues in family psychology research. Applications of confirmatory factor analysis are also discussed as useful methods for researchers examining multiple components of families and handling multiple perspectives of various family members. Limitations and applications of these methods in family psychology research are reviewed.

psychology, clinical,family studies

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are how to handle complex data structures and the selection and application of multivariate analysis methods in family psychology research. Specifically, the authors explored how to use multivariate statistical methods (such as multivariate analysis of variance (MANOVA), discriminant analysis, and confirmatory factor analysis) to solve the following types of problems in family psychology research:

1. **Handling unit - of - analysis problems**: A common problem in family psychology research is determining whether the basic unit of analysis is the individual or the family as a whole. For example, if the hostility of fathers, mothers, and children is measured separately in a family, these scores may be related to each other. If the individual is taken as the unit of analysis, the independence assumption of the general linear model is violated. To avoid this problem, the family can be taken as the unit of analysis, and the scores of each family member can be regarded as dependent variables.

2. **Evaluating mean differences and relationships**: The relationship systems and interactions within the family require analysis methods that can simultaneously evaluate mean differences as well as relationships and processes within the family. For example, MANOVA can be used to simultaneously evaluate mean differences on multiple dependent variables, not just a single variable.

3. **Theory - driven hypothesis testing**: These multivariate statistical methods are not only suitable for testing and exploring mean differences between groups, but are also particularly powerful for testing or developing hypotheses about relationships between variables based on theory. For example, confirmatory factor analysis can help researchers deal with multiple components of the family and handle multiple perspectives of different family members.

4. **Improving research power**: Designs that include multiple dependent variables also affect the Type II error rate (i.e., power) of the research. Multivariate analysis methods can, in some cases, more effectively detect significant differences between groups than single - variable methods, especially when there is a negative correlation between the dependent variables.

5. **Interpreting latent dimensions**: Through discriminant analysis, the linear combinations that best distinguish different groups can be found, thereby revealing the latent dimensions in the data. This is very useful for understanding the internal connections between family relationship variables.

In summary, this paper aims to demonstrate the application of multivariate statistical methods in family psychology research in order to better understand and explain the complex relationships and interaction patterns within the family system.