MODEL PROBLEMS line is not constant. NOTE: the coefficient (~) .remains an unbiased Multicollineariw - a problem in which estimatornonetheless. two independent variables in a model are highly PROBLEMS caused by Heteroscedasticity: related. Multicollinearity is a sample 1} The coefficient (lJ) is not "efficient" or the phenomenon, not an estimation problem. "best" estimator. NOTE: If all of the independentvariablesare checked, 2} The F-test is no longer reliable. and the t-tests are significant, then multicollinearity 3} The t-tests are no longer reliable. shouldnot be a problem. CHECKINGfor Heteroscedasticity: PROBLEMS caused by Multicollinearity: 1} Check the scatterplot of residuals for fan 1} The estimated' coefficient (lJ) may no longer shaped patterns. be an unbiased estimators. 2} Goldfeld-Quandt test 2) Because of the strong interrelationships 3} Breusch-Pagan test between the independent variables, it is difficult 4} White test to disentangle their separate effects on the dependent variable without violating the cetelis POSSIBLE SOLUTIONS to Heteroscedasticity: paribus assumption. . 1} Generalized Least Squares 3) Significant variables. may appear to be 2} Weighted Least Squares insignificant. (i.e., a Type II error) 3} Use weighting via generalized differencing if the true variance is known. CHECKINGfor Multicollinearity: 4} Perform a transformation of the model from f} High R2 with few orno significant t-ratios. linear to log-linear form. 2} Variance Inflation Factor (V1F) Autocorrelation _ a problem in which If V1F> 10, then suspect a problem. = V1F 1 is IDEAL!!! .the residuals are not independent. This 3A} & 3B} should be performed jointly. problem is often. seen in time-series data. It 3A) Analysis of Structure (SAS Procedure) may be the result of an omitted relevant. 3B} Condition Index (Q) variable" or an incorrect functional form. If CI > 30, then suspect a problem. NOTE: may be positivelyor negativelyrelated. NOTE: the coefficient (~) remains an unbiased 4} Variance Proportion estimatornonetheless. If two or more Variance Proportions> 0.5 in a single row, then suspect a problem. caused by Autocorrelation: 5} High pairwise correlations among 1} The variance of the coefficient (lJ) may be independent variables. understated. 2} The F-test is no longer reliable. POSSIBLE SOLUTIONS to Multicollinearity: 3} The t-tests are no longer reliable. 1} Dropping the redundant independent for Autocorrelation: variable; however, be careful not to omit a 1} Check the scatterplot of residuals for signs of relevant var,'able." <See Autocorrelatl'on> patterns. (positive or negative autocorrelation) 2} Combining highly correlated variables 2} Durbin-Watson statistic (NOTE: use of interaction variables.) 3) Using ratios on a first difference; however this to Autocorrelation: may introduce a Heteroscedasticity problem. . 1} Generalized differencing (if p is known) 4} Using ridge regression. 2) The Cochran-Orcutt procedure 5} Using principle component analysis. 3) The Hildreth-Lu procedure 6) Getting more data. 4) Durbin's H test - USE only when there is a 7) Do Nothing! Live with the lesser of two evils. * lagged dependent variable. <See Autocorrelation definition for warning> @ Roger E. Wehr 1999