Publication date: 07/08/2024

Model Comparisons

The Model Comparisons section of the Fatigue Model report appears after you have fit at least one model. This section contains information about the data scaling that is used for calculating the model comparison statistics, as well as a table of the fitted models.

Note: The model comparison statistics, -LogLikelihood and AIC, are based on scaled data. This is the only place in the Fatigue Model where reported statistics are based on scaled data. Using scaled data does not affect model comparisons, so it is appropriate to use scaled data in this section. For more information about the data scaling, see Meeker et al. (2022).

The model comparison table contains the following columns:

Model Specified for

The category of the model. The Box-Cox Loglinear Sigma and Basquin (Inverse Power) models specify the fatigue-life distribution and then derive the fatigue-strength distribution; they are designated as Fatigue Life in this column. The remaining models specify the fatigue-strength distribution and then derive the fatigue-life distribution; they are designated as Fatigue Strength in this column.

Model Relationship

The type of model.

Model Distribution

The distribution used in the model.

N Parm

The number of parameters in the model.

-LogLikelihood

The negative log likelihood value for the model. Smaller values indicate a better fit. See Likelihood, AICc, and BIC in Fitting Linear Models.

AIC

The Akaike Information Criterion (AIC) value for the model. AIC is computed as 2LogLikelihood + 2k, where k is the number of fitted parameters in the model. Smaller values indicate a better fit.

Model Comparisons Report Options

There are three buttons located above the table that enable you to remove selected models, undo the last action, or redo the last undone action.

The Model Comparisons red triangle menu contains the following option:

Close All Individual Reports

Closes the outline nodes for all of the fitted model reports.

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