Limitations and Assumptions of Multivariate Analysis
Although each individual method of multivariate analysis has its own assumptions (discussed at the relevant point in the text), there is one assumption that is common to all, and that is the assumption of linearity.
The assumption is that the outcome changes linearly with each predictor variable. If the predictor variable is linear, then the assumption is that for a linear change in the predictor variable there will be a linear change in the outcome. When the predictor is ordinal, the size of the change in the outcome is the same for each unit change in the predictor.
What will likely change is the scale of the change, depending on the predictor variable. When variable A has a greater effect on the outcome than variable B, then a one-unit change in A will lead to a greater change in the outcome than a one-unit change in B. Each predictor variable may be weighted differently (the coefficients), but the assumption of linearity remains the...
