A comparative analysis of graphical interaction and logistic regression modelling: self-care and coping with a chronic illness in later life

Quantitative research especially in the social, but also in the
biological sciences has been limited by the availability and
applicability of analytic techniques that elaborate interactions
among behaviours, treatment effects, and mediating variables. This
gap has been filled by a newly developed statistical technique,
known as graphical interaction modelling. The merit of graphical
models for analyzing highly structured data is explored in this
paper by an empirical study on coping with a chronic condition as a
function of interrelationships between three sets of factors. These
include background factors, illness context factors and four
self--care practices. Based on a graphical chain model, the direct
and indirect dependencies are revealed and discussed in comparison
to the results obtained from a simple logistic regression model
ignoring possible interaction effects. Both techniques are
introduced from a more tutorial point of view instead of going far
into technical details.