Fitting Linear Models > Statistical Details
Publication date: 08/13/2020

Statistical Details

Fitting Linear Models

This appendix discusses the different types of response models, their factors, their design coding, and parameterization. It also includes many other details of methods described in the main text.

The JMP system fits linear models to three different types of response models that are labeled continuous, ordinal, and nominal. Many details about the factor side are the same for the different response models, but JMP supports graphics and marginal profiles only for continuous responses—not for ordinal and nominal.

Different computer programs use different design-matrix codings, and thus parameterizations, to fit effects and construct hypothesis tests. JMP uses a different coding than the GLM procedure in the SAS system, although in most cases JMP and SAS GLM procedure produce the same results. The following sections describe the details of JMP coding and highlight those cases when it differs from that of the SAS GLM procedure.

Contents

The Response Models

Continuous Responses
Nominal Responses
Ordinal Responses

The Factor Models

Continuous Factors
Nominal Factors
Ordinal Factors

Frequencies

The Usual Assumptions

Assumed Model
Relative Significance
Multiple Inferences
Validity Assessment
Alternative Methods

Key Statistical Concepts

Uncertainty, a Unifying Concept
The Two Basic Fitting Machines

Likelihood, AICc, and BIC

Power Calculations

Computations for the LSN
Computations for the LSV
Computations for the Power
Computations for the Adjusted Power

Inverse Prediction with Confidence Limits

Platforms That Support Validation

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