Multicollinearity and misleading statistical results

Abstract

Multicollinearity represents a high degree of linear intercorrelation between explanatory variables in a multiple regression model and leads to incorrect results of regression analyses. Diagnostic tools of multicollinearity include the variance inflation factor (VIF), condition index and condition number, and variance decomposition proportion (VDP). The multicollinearity can be expressed by the coefficient of determination (Rh2) of a multiple regression model with one explanatory variable (Xh) as the model’s response variable and the others (Xi [i≠h] as its explanatory variables. The variance (σh2) of the regression coefficients constituting the final regression model are proportional to the VIF (11-Rh2). Hence, an increase in Rh2 (strong multicollinearity) increases σh2. The larger σh2 produces unreliable probability values and confidence intervals of the regression coefficients. The square root of the ratio of the maximum eigenvalue to each eigenvalue from the correlation matrix of standardized explanatory variables is referred to as the condition index. The condition number is the maximum condition index. Multicollinearity is present when the VIF is higher than 5 to 10 or the condition indices are higher than 10 to 30. However, they cannot indicate multicollinear explanatory variables. VDPs obtained from the eigenvectors can identify the multicollinear variables by showing the extent of the inflation of σh2 according to each condition index. When two or more VDPs, which correspond to a common condition index higher than 10 to 30, are higher than 0.8 to 0.9, their associated explanatory variables are multicollinear. Excluding multicollinear explanatory variables leads to statistically stable multiple regression models.

Be Aware Not Reactive: Testing a Mediated-Moderation Model of Dark Triad and Perceived Victimization via Self-Regulatory Approach

“…Multicollinearity was also tested through VIF values. All the values were found to be less than 5, thereby confirming that multicollinearity was not an issue in the model (Kim, 2019). Table 2 shows the square root values of AVE on the diagonal (in bold), which confirm the discriminant validity according to the widely used criterion of Fornell-Larcker.…”

COVID-19 worry, mental health indicators, and preparedness for future care needs across the adult lifespan

“…The remaining 485 cases (97%) have complete responses for the variables used in correlational and regression analyses. Problematic mulitcollinearity was not present among the predictors; the largest variance inflation factor (VIF) was equal to 3.2 (for general anxiety), which is below values that are considered to indicate troublesome (VIF > 5, Kim, 2019) or excessive (VIF > 10, Cohen, Cohen, West, & Aiken, 2003) multicollinearity. With the sample size of 485, the regression analyses had ample statistical power to detect effects.…”

Participation in Disease Management Programs and Major Adverse Cardiac Events in Patients After Acute Myocardial Infarction.

“…In order to identify determinants of DMP and CR, we obtained multivariable relative risks from Poisson model with robust error variance as alternative to logistic regression analysis for frequent outcomes [30]. Since investigated determinants were correlated, we formally tested for multicollinearity to exclude erroneous results [31]. We analyzed the effects of CR or DMP participation on the occurrence of MACE using Cox proportional hazard regression models.…”