*Creating models, running regressions, and making predictions*

About models

Models describe relationships in a table and can be used to make predictions

Viewing a model’s adjusted R-squared

The adjusted R-squared is a measure of model fit

Viewing the confidence intervals of a model’s explanatory variables

Values inside a coefficient’s confidence interval are not significantly different from the estimate

Exporting a model as R code

Choose Model > Show Estimation Command to see R code for the current model

Exporting a model as a Stata, SPSS, or SAS command (Pro only)

Choose Model > Show Estimation Command to see command syntax for the current model

Copying a model’s predicted values to the host table

Choose Model > Copy Predictions to Host Table to create a numeric column out of the model’s predicted values

Copying model residuals to the host table

Choose Model > Copy Residuals to Host Table to create a numeric column out of the model residuals

About proportional hazards models

Proportional hazards models can be used to model failure times or life events

Defining the cause of failure in a proportional hazards models

Modeling multiple causes of failure produces separate coefficients for each cause

Treating records as censored in a proportional hazards models

Censoring is appropriate if some records never experience a failure

Defining default outcome values in a proportional hazards models

A default column is useful when censored observations appear in a different column than non-censored observations

Stratifying a proportional hazards model by a category column

Stratification is appropriate when different groups of observations have different underlying hazard rates

Creating a model

Create a multivariate model from the Summary view

Deleting a model

A model can be deleted from the Model menu

Duplicating a model

Duplicate a model by control-clicking it in the navigator

Adding and removing variables from a model

A model’s list of explanatory variables can be modified after creation

Treating an explanatory variable as categorical data

Categorical explanatory variables are appropriate for discrete data

Exporting a model as an interactive spreadsheet

Export a model’s coefficients and predictions as an interactive Excel spreadsheet

Exporting a model’s coefficient table

Export a model’s coefficient table as CSV, Excel, or JSON for use in other programs

Including higher-order terms of a numeric explanatory variable in a model

Higher-order terms can provide a better model fit, but should be used with caution

Creating an interaction term in a model

Interaction terms control for the interaction effects of two explanatory variables

Testing hypotheses about multiple coefficients in a model

Test model coefficients for equality, or whether they sum to 0 or 1

Testing the joint significance of multiple coefficients in a model

Perform a combined test by selecting multiple rows in the explanatory variable table

Viewing a model’s log-likelihood

The log-likelihood is a measure of model fit

Viewing the odds ratios of a model

The odds ratio is an alternative representation of a model’s coefficients

Specifying the omitted category of a categorical explanatory variable

When an explanatory variable is treated as categories, its estimated coefficients are relative to an omitted category

Setting the base outcome of a categorical outcome variable

When the outcome variable is categorical, estimated coefficients are always relative to a base outcome

Treating an outcome variable as categorical data

Categorical outcome variables are appropriate for discrete-response models

Treating an outcome variable as continuous data

Treat an outcome as continuous if it takes a range of ordered values

Setting the count exposure of an outcome variable

An exposure column defines the relative opportunity that each observation had to receive events

Viewing the p-values of a model’s explanatory variables

The p-value conveys the statistical significance of an model coefficient

Viewing a model’s unadjusted R-squared

The unadjusted R-squared is a measure of model fit

Renaming a model

A model can be renamed by double-clicking it in the navigator

Rearranging models

A model can be rearranged via drag-and-drop in the navigator

Viewing the z-scores or t-statistics of a model’s explanatory variables

The z-score or t-statistic equals the coefficient divided by the standard error, and reflects coefficient’s statistical significance