1.0 - 8.00 - GLM Onscreen Output - Teradata Vantage

Teradata® Vantage Machine Learning Engine Analytic Function Reference

Teradata Vantage
Release Number
Release Date
May 2019
Content Type
Programming Reference
Publication ID
English (United States)
The onscreen output of the GLM function is a table containing information about the regression analysis of the data, in two sections:
  • Information about the model intercept and coefficients
  • Information about the regression (number of iterations and number of rows processed) and several goodness-of-fit measures


Column Description

Name of predictor or other reported result.

For categorical predictors, the function selects one category as the reference category, and outputs one row for each other category for the column, in the format predictor.level.

For example, if column color has categories 'red', 'blue', and 'green', and green is the reference category, the function outputs these rows:



estimate For predictors, estimated value of coefficient.

For other reported results, calculated value.

std_error For predictors, standard deviation of the mean (standard error).

For other reported results, not applicable (value 0).

z_score For predictors, calculated z-score.

For other reported results, calculated value.

p_value For predictors, calculated p-value.

For other reported results, not applicable (value 0).

significance For predictors, indicator of predictor significance. For key to significance codes, see CoxPH Output.

For other reported results, description of result.


The onscreen output includes a row for each estimated parameter of the model and additional information about the model and the regression.

For Step ('true'), the rows of the following tables appear for the model at each step and describe the model at that step. For each step, the description includes only the predictors included in the model.

Estimated Parameters of the Model
Parameter Description
(Intercept) Value of link function (Y) when all predictors are 0.
predictor [Appears only for numerical predictor.] Predictor name.
predictor.level [Appears only for categorical predictor.] Predictor name and level. Table has a row for each level of the predictor except one, which serves as the reference level.
Model and Regression Information
Value Description
ITERATIONS# Number of Fisher Scoring iterations performed on function.
With Step ('true'), function reports this number for each step.
ROWS# Number of rows of data received as input.
Residual deviance Deviance, with degrees of freedom reported in significance column.
Residual deviance is not displayed when Family is GAMMA, NEGATIVE_BINOMIAL, or INVERSE_GAUSSIAN.
Pearson goodness of fit Sum of squared Pearson residual.
AIC Akaike information criterion, a measure of relative quality of model for given data set.
BIC Bayesian information criterion, partly based on likelihood function and closely related to AIC. BIC is a criterion for model selection among a finite set of models; the model with the lowest BIC is preferred.
Wald Test Tests goodness of fit.
Dispersion parameter For GAUSSIAN, the value of this parameter is estimated from the data. For all other families, this parameter value is 1.

The coefficients are also stored in the table output_table for later use.

For the Gamma distribution density, AIC and BIC might have the value NaN when the dispersion parameter is very small and goodness-of-fit is poor.