Check predictor overlap with variance inflation diagnostics fast. Review correlations, tolerance, and interpretation guidance easily. Export clean tables and visualize relationships for stronger decisions.
| X1 | X2 | X3 | X4 |
|---|---|---|---|
| 10 | 12 | 9 | 4 |
| 12 | 14 | 10 | 5 |
| 14 | 17 | 12 | 7 |
| 16 | 19 | 13 | 8 |
| 18 | 22 | 15 | 9 |
| 20 | 24 | 16 | 10 |
This sample intentionally contains related predictors. It helps test VIF, tolerance, determinant, and condition index behavior before you paste your own dataset.
Multicollinearity appears when predictor variables move together too strongly. That overlap makes regression coefficients unstable. Standard errors grow. Signs can flip. Small data changes may create large model changes. This calculator helps you inspect that overlap before you trust a model.
The tool reads a numeric predictor dataset and builds the correlation matrix. It then estimates variance inflation factor, tolerance, determinant, eigenvalues, and condition indices. These measures show whether one predictor can be explained by the others. When that happens, your model may still predict well, but interpretation becomes weaker.
Use this page during feature screening, model review, or classroom analysis. It is useful for statistics, econometrics, machine learning preparation, and general mathematics work. The visual plots help you see whether the problem comes from many mild relationships or a few very strong ones.
Variance Inflation Factor: VIFj = 1 / (1 - Rj2)
Tolerance: Tolerancej = 1 / VIFj = 1 - Rj2
Correlation matrix determinant: a very small determinant suggests strong linear dependence among predictors.
Condition Index: CIi = sqrt(λmax / λi)
Here, Rj2 comes from regressing one predictor on all remaining predictors, and λ values are eigenvalues from the correlation matrix. Higher VIF and higher condition index usually signal stronger multicollinearity. Very low tolerance and a near zero determinant point in the same direction.
A common rule is that VIF above 5 deserves review and VIF above 10 suggests a serious issue. Pairwise correlation above 0.80 is also a useful warning, but it should not replace the full VIF check.
It means two or more predictors share overlapping information. The model struggles to separate their individual effects, so regression coefficients become less stable and harder to interpret.
Many analysts review variables when VIF exceeds 5. Values above 10 often indicate a stronger problem. The exact cutoff depends on your field, sample size, and modeling goal.
Tolerance is the inverse of VIF. Small tolerance means one predictor is largely explained by other predictors. It provides the same warning in a different scale.
No. Pairwise correlation can miss cases where several variables together create dependency. VIF and condition index can reveal issues that simple one to one correlations may hide.
A very small determinant suggests the correlation matrix is close to singular. That usually means strong linear dependence among predictors and unstable coefficient estimates.
They summarize how stretched the predictor space becomes. Larger condition indices suggest stronger dependence patterns. They are especially helpful when several variables contribute to the issue together.
Not always. You might combine variables, center them, collect more data, or keep them for prediction. Removal depends on whether interpretation or raw predictive accuracy matters more.
Yes. It is useful before linear regression, logistic regression, and feature engineering. It helps you screen predictors before training or explaining a model.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.