Today, the linear model of regression has been generalized (hence the general linear model, GLM, used in multinomial logistic regression) so that SPSS and other statistical packages can create dummy variables automatically for all categorical inputs and can generate computer coefficients (parameters) for variables regardless of level. Moreover, new estimation techniques, particularly maximum likelihood estimation (MLE), have largely replaced the older ordinary least squares (OLS) forms of estimation on which regression was based. The result is that level of data is less of a restriction today than it once was, ifyou are using new statistical procedures such as GLM multiple analysis of variance (MANOVA), multinomial logistic regression, or structural equation modeling. Also, SPSS CATEGORIES module supports a wide range of procedures for categorical data: PRINCALS, OVERALS, and CATREG are analogous to principal components, canonical correlation, and regression.