Prediction Error in Multiple Regression Model Cover Image

Prediction Error in Multiple Regression Model
Prediction Error in Multiple Regression Model

Author(s): Slav Angelov, Eugenia Stoimenova
Subject(s): Economy, ICT Information and Communications Technologies
Published by: Нов български университет
Keywords: Prediction error; Multicollinearity; Regression Analysis; RMSECV; Crossvalidation

Summary/Abstract: In this paper will be briefly proposed a regression technique for reducing the prediction error (RMSECV) in multiple regression model. It is based on a method for estimating the model’s regressors based on leave-one-out crossvalidation. Two variations of the method are proposed - one using the Root mean square error (RMSE) and one using the mean absolute error (MAE). Using R language the technique is tested on a real problem concerning financial ratios and the results are compared with the OLS regression error. For the presented example both variations of the method are with lower RMSECV than the OLS one. For the presented example it is observed that when one of the two variations of the proposed method does not reduce the OLS error of an observation than the other does it. Possible advantages when this event occurs are discussed.

  • Issue Year: 13/2017
  • Issue No: 1
  • Page Range: 329-347
  • Page Count: 19
  • Language: English