Clarifying gender interactions in multivariate analysis

To examine a linear regression model to predict physical fitness using an alternative concept to analyse gender interactions.

Data were obtained from the Berne Lifestyle Panel, a survey on health and lifestyles of 56-66 years old Bernese citizens. A measure of physical fitness was regressed on gender, education and their interaction as central explanatory variables and age as confounding factor. For ease of interpretation, two dummy variables of education are introduced, one for female, the other for male education. The model with education dummy variables is compared to a linear regression model without interaction stratified by gender, and with a model with the multiplicative gender-education interaction term without stratification.

The use of dummy variables ensures an accurate description of both women's and men's associations of education with the dependant variable, without losing any explanatory power due to stratification.

The results show that the use of dummy variables is a rewarding alternative to stratification and conventional gender interaction analysis, providing both sufficient statistical information and a basis for straightforward interpretation.