Skip to content

Latest commit

 

History

History
51 lines (42 loc) · 1.46 KB

README.md

File metadata and controls

51 lines (42 loc) · 1.46 KB
Raw

Determining Top Predictors for ISLR Hitters Dataset

LASSO, elastic net, Adaptive LASSO, SCAD methods for determining top predictors for each method.

For Lasso below are the top predictors:
  • Hits 1.72206743
  • Walks 1.57275905
  • CHmRun 0.49835782
  • CRuns 0.08740234
  • CRBI 0.50979015
  • PutOuts 0.18776039
For Elastic Net below are the top predictors:
  • Hits 1.58604917
  • Walks 1.63296070
  • CHits 0.03275671
  • CHmRun 0.90455877
  • CRuns 0.16208771
  • CRBI 0.25834082
  • DivisionW -1.52527575
  • PutOuts 0.17776448
For SCAD below are the top predictors:
  • AtBat -2.1672572
  • Hits 8.1885705
  • HmRun -8.8484358
  • Walks 6.0337992
  • Years -10.2448385
  • CHits 0.3680495
  • CHmRun 2.9314344
  • CWalks -0.5432212
  • DivisionW -80.0966560
  • PutOuts 0.3353118
  • Assists 0.1743041
  • NewLeagueN -3.5558069
For Adaptive LASSO below are the top predictors:
  • Hits 2.3337452
  • Walks 3.3849602
  • CRBI 0.4805223

Hits predictor is common in all methods whoever each method emphasizes different coefficient values. All the methods accept SCAD positive correlation impact on predicting the response variable Salary.

If we look at the pair graphs amongst these variables: CRBI have comparatively more correlation with response variable Salary. Also, AtBat and Hits are highly correlated.

Even though, SCAD does not have CRBI as one of its top predictors.