ANALISIS AUTOKORELASI PADA MODEL REGRESI LINIER BERGANDA
| Main Author: | NOVITA SARI, AYU |
|---|---|
| Format: | Thesis NonPeerReviewed |
| Terbitan: |
, 2013
|
| Subjects: | |
| Online Access: |
http://eprints.umm.ac.id/15740/ |
Daftar Isi:
- Analysis of the multiple linear regression model is used to predict dependent variable Y with more than one independent variable X. Predicted value of variable Y, variable X must be determined and a as regression parameters are known. One way to determine the value of the regression parameters is through an estimation by the ordinary least squares (OLS) method. The principle of this method is minimize the sum of squared errors based on observations of actual value Y to observations of estimates value Y. OLS method has assumptions that must fulfilled for the estimation regression parameters which is obtained has properties BLUE (Best Linear unbiased Estimator). This study is talk about one of irregularities in the OLS assumptions of multiple linear regression model, namely auto correlation which a disturbed error u_i are not independent (correlated) to other disturbed errors (u_j) with E(u_i u_j )≠0 for i≠j value. Therefore, this study describes the influence of auto correlation on regression model and describes how to cope model with auto correlation if has serious impact on regression model. The method in this study is use literature study which is relevant to the issues to do an analysis of the correctness and suitability between theory and the problems so that can provide the appropriate conclusions. The influence of auto correlation that occurs in multiple linear regression model causes OLS estimator is not BLUE so auto correlation should be corrected. Prior to correct, the model must be detected about the presence or absence of auto correlation by graph method, statistic d Durbin Watson and run test. Based on the detection, if there is evidence of auto correlation it should be corrected, by the steps (1) Determine the estimation value of the auto correlation coefficient γ (by way of (a) estimation method γ based on the statistical d Durbin Watson, (b) estimation methods γ based statistic d by Theil and Nagar, (c) estimation method γ Cochrane Orcutt, (d) two-step method estimation γ Durbin) (2) Do the transformation on the original data by using estimation γ (3) Applying OLS on the data transformed to obtain multiple linear regression model that is free auto correlation.