Desiderata for a statistical estimator
Before discussing the assumptions of multiple
regression, it is important to discuss what we need to
make these assumptions for. Remembering that a
regression coefficient based on sample data is an
estimate of a true regression parameter for the
population the sample is drawn from, there are three
particularly important properties for a statistical
estimator (Dougherty, 2007). These three properties are
true of regression coefficients calculated via ordinary
least squares—provided that certain assumptions are
met (Cohen et al., 2003).