View Image

Despite the fact that many phenomenon in physical or social sciences involve interaction between causal factors, it seems that very few researchers using multiple regression or structural equation modeling techniques do investigate for the presence of those interactions. If fact, in the context of multiple regression, a statistically significant interaction is often perceived as a curse by researchers, as something that jeopardize their effort to identify simple causal factors.

This perception is somewhat reinforced by the fact that many statistical textbooks refer to a statistically significant interaction solely as an indication that an important assumption of multiple regression has been violated (namely the additivity assumption) without further instruction on how to deal with such an interaction. Some of those textbooks even suggest that the net result of such a significant interaction is that the regression model become uninterpretable. Nothing is farther from the truth!

Part of this misconception is caused by a lack of understanding of what interaction is all about and by an unfamiliarity with how interaction should be interpreted. While it is perfectly possible to deduce the nature of the interaction simply by examining the resulting regression equation, a common way to get an intuitive feel for the interaction is to compute several regression slopes relating one of the independent variable to the dependent variable at different values of the other independent variable. However, this task is cumbersome especially when investigating several possible interactions.

The ITALASSI application was developed to be a free program that has been written to facilitate interpretation of regression models (2 independent variables) with an interaction term.

The program allows you to enter several regression models (two bivariate, one multiple additive, and one multivariate with interaction) in the form of equations or compute those equations from raw data and displays the various models using 2D and 3D graphs. The program may also be used in advanced stat courses to illustrate statistical interactions or applied multiple regression.