1.0 - 8.00 - LAR 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 function outputs a message and a model table.

Output Message Schema

Column Data Type Description
message VARCHAR Reports that the result is stored in the table specified in the OutputTable argument.

OutputTable Schema

This is the model table to input to LARPredict.

Column Data Type Description
steps INTEGER Sequence number of step. One LAR or LASSO move represents one step.
var_id INTEGER Sequence number of predictor. Sequence of predictors is specified by TargetColumns argument.
var_name VARCHAR Column name of predictor.
max_abs_corr DOUBLE PRECISION Modified maximum absolute correlation (common for all active variables) between active variables and current residuals. This value is not necessarily in the range [0,1].
step_length DOUBLE PRECISION Distance to move along equiangular direction in step.
intercept DOUBLE PRECISION Constant item in model. Value evolves along path.
predictor_column DOUBLE PRECISION [Column appears once for each predictor_column.] Coefficient for predictor.

Interpreting the Output

At the beginning of stepi, the variable Xk (identified by the values in columns var_id and var_name) either enters into the regression model (indicated by a positive value in the column var_id) or drops from the regression model (indicated by a negative value in the column var_id), and the current common correlation between active variables and current residuals is the value in the column max_abs_corr.

After moving along the equiangular direction for the distance in the column step_length, either an inactive variable qualifies to enter into the model or a currently active variable is dropped from the model, whereby the process reaches stepi+1. The intercept and coefficients correspond to both the end of stepi and the beginning of stepi+1.