This example uses the same elastic net and Gaussian regression as GLM2 Example 3: Elastic Net for Gaussian Regression Analysis, but instead of specifying the number of lambdas to use, it specifies a set of specific lambda values. Teradata recommends this practice when the you have strong prior knowledge of the magnitude of lambda or want to test the results of predetermined lambda values.
Input
The input table is glm2_elastic_net, as in GLM2 Example 3: Elastic Net for Gaussian Regression Analysis.
SQL Call
The function calculates predictions only for the three specified lambda values.
SELECT * FROM GLM2 (
ON glm2_elastic_net AS InputTable
OUT TABLE ModelTable (glm2_elastic_net_2model)
OUT TABLE RegularizationTable (glm2_4regularization)
USING
InputColumns ('gnp_deflator','gnp','armed_forces','population')
ResponseColumn ('employed')
Family ('GAUSSIAN')
Intercept ('TRUE')
Lambda (0.03,0.05,0.08)
Alpha(0.1)
) AS dt;
Output
dfDevRatio | devRatio | dfDev | deviance | lambda |
---|---|---|---|---|
4 | 0.962764132079392 | 11 | 6.88896420908282 | 0.08 |
3 | 0.965249423725026 | 12 | 6.42916331945636 | 0.05 |
4 | 0.969563681197274 | 11 | 5.63098760945407 | 0.03 |
df_dev_ratio | deviance_ratio | df_dev | deviance | lambda | intercept | armed_forces | gnp | gnp_deflator | population |
---|---|---|---|---|---|---|---|---|---|
4 | 0.962764132079392 | 11 | 6.88896420908282 | 0.08 | 44.5464394715 | 0.0013604856 | 0.0232838328 | 0.0786912064 | 0.0288475833 |
3 | 0.965249423725026 | 12 | 6.42916331945636 | 0.05 | 48.2203770905 | 0.0011161847 | 0.0274548885 | 0.0605957447 | 0 |
4 | 0.969563681197274 | 11 | 5.63098760945407 | 0.03 | 55.8031821863 | 0.0005469725 | 0.0351076729 | 0.0348849624 | -0.0663160626 |
The function also outputs a model table, which is not shown here.