# Statistical Advisor, General Nonlinear Regression

GENERALIZED LINEAR MODEL (GLZ). This chapter discusses an implementation of the generalized linear model and how to compute a standard, stepwise, or best subset multiple regression analysis, in which one (non-normally distributed) dependent (criterion, endogenous) variable is linearly or non-linearly (see Link Function) related to multiple independent (predictor, exogenous) variables.

NONLINEAR ESTIMATION. This chapter discusses techniques for estimating any type of regression equation. In most general terms, all multiple regression equations, that is, relationships between variables, can be expressed as y = F(x). In other words, the dependent variable y is a function of one or more variables x. The nature of the function may be linear, exponential, logarithmic, or even discontinuous.

For example, we could explicitly specify a so-called breakpoint regression model. That is, we may hypothesize that the relationship between x and y is generally linear, but that the slope is different in different regions of the x variable. Thus, the number of bedrooms in a house may be related to the price of the house; however, the relationship may not be as strong for large houses that have more than 5 bedrooms.

NONLINEAR ESTIMATION allows you to type in any kind of regression equation, and to specify how it is to be estimated (least squares, weighted least squares, etc.)

GRAPHICAL ANALYTIC TECHNIQUES: Graphical analytic techniues include various facilities for 2D and 3D fitting of lines or surfaces to observed data. These graphs offer a wide variety of methods to visualize relationships between variables.