I have the following pipeline.
First I apply ridge regression using 10-cv to find the best lambda.
I get same lambda max and lambda best as in R cv.glmnet.
Next, I refit the model using the best lambda from the first step, without intercept and compare it to the results of R glmnet.
The coefficients and predictions are different. Why is that?
Comparison of coefficients:
R
(Intercept) 0
f1 -0.004059542
f2 0.377331808
f3 1.006589044
f4 0.876858914
f5 0.140710854
f6 730268.470575249
f7 244447.850561236
f8 537663.923355049
f9 176279.892636801
f10 662.748853227
f11 739399.127039033
Hi,
I have the following pipeline. First I apply ridge regression using 10-cv to find the best lambda. I get same lambda max and lambda best as in R cv.glmnet.
Next, I refit the model using the best lambda from the first step, without intercept and compare it to the results of R glmnet. The coefficients and predictions are different. Why is that?
Comparison of coefficients: R (Intercept) 0
f1 -0.004059542 f2 0.377331808 f3 1.006589044 f4 0.876858914 f5 0.140710854 f6 730268.470575249 f7 244447.850561236 f8 537663.923355049 f9 176279.892636801 f10 662.748853227 f11 739399.127039033
python: Intercept 0 f1 -0.16957 f2 0.33352 f3 0.80749 f4 0.71330 f5 0.11385 f6 801091.27661 f7 293769.02256 f8 557147.70998 f9 251954.31707 f10 797640.12411 f11 1086129.27954
Thanks