For the latest version of JMP Help, visit JMP.com/help.


Fitting Linear Models > Generalized Linear Models > Statistical Details for the Generalized Linear Model Personality
Publication date: 11/29/2021

Statistical Details for the Generalized Linear Model Personality

To construct a generalized linear model, you must select response and explanatory variables for your data. You then must choose an appropriate link function and probability distribution for your response. Explanatory variables can be any combination of continuous variables, classification variables, and interactions. Some common examples of generalized linear models are listed in Table 12.1.

Table 12.1 Examples of Generalized Linear Models

Model

Response Variable

Distribution

Default Link Function

Traditional Linear Model

continuous

Normal

identity, g(μ) = μ

Logistic Regression

a count or a binary random variable

Binomial

logit,Equation shown here

Poisson Regression in Log Linear Model

a count

Poisson

log, g(μ) = log(μ)

Exponential Regression

positive continuous

Exponential

Equation shown here

The platform fits a generalized linear model to the data by maximum likelihood estimation of the parameter vector. In general, there is no closed-form solution for the maximum likelihood estimates of the parameters. Therefore, the platform estimates the parameters of the model numerically through an iterative fitting process using a technique pioneered by Nelder and Wedderburn (1972). The overdispersion parameter φ is estimated by dividing the Pearson goodness-of-fit statistic by its degrees of freedom. Covariances, standard errors, and confidence limits are computed for the estimated parameters based on the asymptotic normality of maximum likelihood estimators.

A number of link functions and probability distributions are available in the Generalized Linear Model personality of the Fit Model platform. Table 12.2 lists the built-in link functions.

Table 12.2 Built-in Link Functions

Link Function Name

Link Function Formula

Identity

g(μ) = μ

Logit

Equation shown here

Probit

g(μ) = Φ-1(μ), where Φ is the standard normal cumulative distribution function

Log

g(μ) = log(μ)

Reciprocal

g(μ) =Equation shown here

Power

Equation shown here

Comp LogLog

g(m) = log(–log(1 – μ))

When you select the Power link function, a number box appears that enables you to enter the desired power.

Table 12.3 lists the variance functions associated with the available distributions for the response variable.

Table 12.3 Variance Functions for Response Distributions

Distribution

Variance Function

Normal

V(μ) = 1

Binomial

V(μ) = μ(1 – μ)

Poisson

V(μ) = μ

Exponential

V(μ) = μ2

Want more information? Have questions? Get answers in the JMP User Community (community.jmp.com).