Variance inflation factor

In statistics, the variance inflation factor (VIF) is the ratio (quotient) of the variance of estimating some parameter in a model that includes multiple other terms (parameters) by the variance of a model constructed using only one term. It quantifies the severity of multicollinearity in an ordinary least squares regression analysis. It provides an index that measures how much the variance (the square of the estimate's standard deviation) of an estimated regression coefficient is increased because of collinearity. Cuthbert Daniel claims to have invented the concept behind the variance inflation factor, but did not come up with the name.

Definition

Consider the following linear model with ''k'' independent variables:
: ''Y'' = ''β''₀ + ''β''₁ ''X''₁ + ''β''₂ ''X'' ₂ + ... + ''β''_''k'' ''X''_''k'' + ''ε''.

The standard error of the estimate of ''β''_''j'' is the square root of the ''j'' + 1 diagonal element of ''s''²(''X''′''X'')⁻¹, where ''s'' is the root mean squared error (RMSE) (note that RMSE² is a consistent estimator of the true variance of the error term, $\sigma^2$); ''X'' is the regression design matrix — a matrix such that ''X''_{''i'', ''j''+1} is the value of the ''j''^th independent variable for the ''i''^th case or observation, and such that ''X''_''i'',1, the predictor vector associated with the intercept term, equals 1 for all ''i''. It turns out that the square of this standard error, the estimated variance of the estimate of ''β''_''j'', can be equivalently expressed as:

:$\widehat(\hat_j) = \frac\cdot \frac,$

where ''R''_''j''² is the multiple ''R''² for the regression of ''X''_''j'' on the other covariates (a regression that does not involve the response variable ''Y''). This identity separates the influences of several distinct factors on the variance of the coefficient estimate:

* ''s''²: greater scatter in the data around the regression surface leads to proportionately more variance in the coefficient estimates
* ''n'': greater sample size results in proportionately less variance in the coefficient estimates
* $\widehat\operatorname(X_j)$: greater variability in a particular covariate leads to proportionately less variance in the corresponding coefficient estimate

The remaining term, 1 / (1 − ''R''_''j''²) is the VIF. It reflects all other factors that influence the uncertainty in the coefficient estimates. The VIF equals 1 when the vector ''X''_''j'' is orthogonal to each column of the design matrix for the regression of ''X''_''j'' on the other covariates. By contrast, the VIF is greater than 1 when the vector ''X''_''j'' is not orthogonal to all columns of the design matrix for the regression of ''X''_''j'' on the other covariates. Finally, note that the VIF is invariant to the scaling of the variables (that is, we could scale each variable ''X''_''j'' by a constant ''c''_''j'' without changing the VIF).

:
\widehat(\hat_j) = s^2 X^T X)^

Now let $r= X^T X$, and without losing generality, we reorder the columns of ''X'' to set the first column to be $X_j$
:$r^ = \begin r_ & r_ \\ r_ & r_\end^$

: $r_ = X_j^T X_j, r_ = X_j^T X_, r_ = X_^T X_j, r_ = X_^T X_$.

By using Schur complement, the element in the first row and first column in $r^$ is,
:r^_ = _ - r_ r_^ r_
Then we have,
:
\begin
& \widehat(\hat_j) = s^2 X^T X)^ = s^2 r^_ \\
= & s^2 _j^T X_j - X_j^T X_ (X_^T X_)^ X_^T X_j \\
= & s^2 _j^T X_j - X_j^T X_ (X_^T X_)^ (X_^T X_) (X_^T X_)^ X_^T X_j \\
= & s^2 _j^T X_j - \hat_^T(X_^T X_) \hat_ \\
= & s^2 \frac \\
= & \frac\cdot \frac
\end

Here $\hat_$ is the coefficient of regression of dependent variable $X_j$ over covariate $X_$. $\mathrm_j$ is the corresponding residual sum of squares.

Calculation and analysis

We can calculate ''k'' different VIFs (one for each ''X''_''i'') in three steps:

Step one

First we run an ordinary least square regression that has ''X''_''i'' as a function of all the other explanatory variables in the first equation.
If ''i'' = 1, for example, equation would be
:$X_1=\alpha_0 + \alpha_2 X_2 + \alpha_3 X_3 + \cdots + \alpha_k X_k +e$

where $\alpha_0$ is a constant and ''e'' is the error term.

Step two

Then, calculate the VIF factor for $\hat\beta_i$ with the following formula :

: $\mathrm_i = \frac$

where ''R''²_''i'' is the coefficient of determination of the regression equation in step one, with $X_i$ on the left hand side, and all other predictor variables (all the other X variables) on the right hand side.

Step three

Analyze the magnitude of multicollinearity by considering the size of the $\operatorname(\hat \beta_i)$. A rule of thumb is that if $\operatorname(\hat \beta_i) > 10$ then multicollinearity is high (a cutoff of 5 is also commonly used). However, there is no value of VIF greater than 0 in which the variance of the slopes of predictors isn't inflated. As a result, including two or more variables in a multiple regression that are not orthogonal (i.e. have correlation = 0), will alter each other's slope, SE of the slope, and P-value, because there is shared variance between the predictors that can't be uniquely attributed to any one of them.

Some software instead calculates the tolerance which is just the reciprocal of the VIF. The choice of which to use is a matter of personal preference.

Interpretation

The square root of the variance inflation factor indicates how much larger the standard error increases compared to if that variable had 0 correlation to other predictor variables in the model.

Example

If the variance inflation factor of a predictor variable were 5.27 (√5.27 = 2.3), this means that the standard error for the coefficient of that predictor variable is 2.3 times larger than if that predictor variable had 0 correlation with the other predictor variables.

Implementation

*vif function in th
car
R package
*ols_vif_tol function in th
olsrr
R package
*PROC REG in SA
System
*variance_inflation_factor function i
statsmodels
Python package
*estat vif i
Stata
addon for GRASS GIS

References

Further reading

*
*
*
*
*
*
* {{cite journal , last1=Zuur , first1=A.F. , last2=Ieno, first2=E.N., last3=Elphick, first3=C.S, year=2010 , title=A protocol for data exploration to avoid common statistical problems , journal=Methods in Ecology and Evolution , volume=1 , pages=3–14 , doi=10.1111/j.2041-210X.2009.00001.x , s2cid=18814132

See also

* Design effect

Regression diagnostics
Statistical ratios
Statistical deviation and dispersion

de:Multikollinearität#Varianzinflationsfaktor