LoadingLoading
Variable inflation factor (VIF) is a measure of the degree of multicollinearity in regression analysis.
Variable inflation factor (VIF) is a measure of the degree of multicollinearity in regression analysis. Multicollinearity occurs when multiple independent variables are correlated in a multiple regression model, which can adversely affect the regression results. Therefore, the variable inflation factor can estimate the degree of inflation of regression coefficient variance due to multicollinearity.
The variable inflation factor is a tool for identifying the degree of multicollinearity. Multiple regression is used when a researcher wishes to test the effect of multiple variables on a specific outcome. The dependent variable is the result of being affected by the independent variable. Multicollinearity is the existence of a linear relationship or correlation between one or more independent variables.
Multicollinearity causes problems in multiple regression models because the independent variables influence each other and are not actually independent, making it difficult to test how combinations of independent variables affect the dependent variable in a regression model.
Although multicollinearity does not reduce the overall predictive power of the model, it may result in estimates of regression coefficients that are not statistically significant. In a sense, this can be viewed as a double-counting phenomenon in the model.
In statistics, multiple regression models with high multicollinearity can make it more difficult to estimate the relationship between each independent variable and the dependent variable. In other words, when two or more independent variables are closely related or measure nearly the same thing, the potential effects they measure are double-counted across the variables. When independent variables are highly correlated, it can be difficult to determine which variable is affecting the dependent variable.
Small changes in the data or adjustments to the structure of the model equations can lead to large and unstable changes in the estimated coefficients of the independent variables. This is a problem because the goal of many econometric models is to accurately test this statistical relationship between independent and dependent variables.
To ensure that the model is appropriately specified and functioning properly, tests for multicollinearity can be performed. The variable inflation factor is one such measurement tool. Utilizing variable inflation factors can help identify the severity of any multicollinearity issues so that model adjustments can be made. The variable inflation factor measures the extent to which the behavior (variance) of an independent variable is affected by the interactions or correlations of other independent variables.
Variable inflation factors provide a quick measure for evaluating the contribution of variables to the standard error in a regression. When significant multicollinearity problems exist, the variable inflation factor for the dependent variables will be very large. Once these variables are identified, several methods can be used to eliminate or merge collinear variables to resolve multicollinearity issues.
The formula of VIF is:
<!-- @@@ -->When
<!-- @@@ -->When equal to 0, the VIF or tolerance is 1, which means that the
<!-- @@@ -->The independent variables are uncorrelated with the remaining independent variables, indicating that there is no multicollinearity.
Generally speaking:
The higher the VIF, the more likely multicollinearity is present and requires further study. When the VIF exceeds 10, there is significant multicollinearity that needs to be corrected.
For example, suppose an economist wishes to test whether there is a statistically significant relationship between the unemployment rate (the independent variable) and the inflation rate (the dependent variable). If other independent variables related to the unemployment rate, such as new initial unemployment claims, are included in the model, multicollinearity may be introduced.
The overall model may show strong statistical explanatory power, but it is impossible to determine whether the effect is primarily due to the unemployment rate or new initial unemployment claims. This is the problem that VIF can detect, and it may suggest removing one of the variables from the model, or finding some way to combine them to capture their joint effect, depending on the hypothesis the researcher wants to test.
As a rule of thumb, VIF values below 3 do not cause concern. As the VIF increases, the reliability of your regression results will decrease.
A VIF equal to 1 means that the independent variables are uncorrelated and there is no multicollinearity in the regression model.
VIF is used to measure the strength of the correlation between independent variables in regression analysis. This correlation is called multicollinearity and can cause problems for regression models.
While moderate multicollinearity is acceptable in regression models, high multicollinearity may be a cause for concern.
Two measures can be taken to correct for high multicollinearity. First, one or more highly correlated variables can be eliminated because the information provided by these variables is redundant. The second approach is to use principal component analysis or partial least squares regression instead of ordinary least squares regression, which can reduce the variables to a smaller uncorrelated set or create new uncorrelated variables, respectively. This will improve the predictability of the model.
[1] CFI. "Variance Inflation Factor."
[2] Isixsigma. "Variance Inflation Factor (VIF)."
TheQuantHackers is a cutting-edge community composed of a group of people from the world's top universities and quantitative finance practitioners, dedicated to exploring the infinite possibilities in the fields of artificial intelligence (AI) and quantification (Quant). Our team members come from world-renowned universities such as the University of Cambridge, Oxford University, Harvard University, ETH Zurich, Peking University, and the University of Science and Technology of China. Our external consultants come from first-class companies such as Microsoft, HSBC, Citadel, Man Group, Citi, Jump Trading, and top domestic private equity companies.
当 变量膨胀因子（VIF）是回归分析中衡量多重共线性程度的一种指标。当多个自变量在多元回归模型中存在相关性时，便会产生多重共线性，这会对回归结果产生不利影响。因此，变量膨胀因子能够估计由于多重共线性导致的回归系数方差的膨胀程度。
变量膨胀因子是识别多重共线性程度的工具。当研究者希望测试多个变量对特定结果的影响时，会使用多元回归。因变量是受到自变量影响的结果。而多重共线性在于一个或多个自变量之间存在线性关系或相关性。
多重共线性在多元回归模型中造成问题，因为自变量之间相互影响，从而实际上并非独立，这使得在回归模型中测试自变量组合如何影响因变量变得困难。
尽管多重共线性不降低模型的整体预测能力，但可能会导致回归系数的估计值不具统计显著性。在某种意义上，这可以被视为模型中一种重复计数的现象。
在统计学中，存在高多重共线性的多元回归模型会使估计出每个自变量与因变量之间的关系变得更加困难。换句话说，当两个或多个自变量紧密相关或几乎测量相同内容时，它们所测量的潜在效应就会在变量之间被重复计算。当自变量之间高度相关时，很难确定哪一个变量在影响因变量。
数据的小幅变化或模型方程结构的调整可能会导致自变量估计系数的巨大和不稳定变化。这是一个问题，因为许多计量经济模型的目标正是准确测试自变量与因变量之间的这种统计关系。
为了确保模型的适当指定和正常运作，可以进行多重共线性测试。变量膨胀因子就是这样的一种衡量工具。利用变量膨胀因子有助于识别任何多重共线性问题的严重性，以便进行模型调整。变量膨胀因子衡量了自变量的行为（方差）受到其他自变量相互作用或相关性的影响程度。
变量膨胀因子为评估回归中变量对标准误差的贡献提供了快速的度量。当存在显著的多重共线性问题时，相关变量的变量膨胀因子将非常大。确定这些变量后，可以使用多种方法来消除或合并共线性变量，以解决多重共线性问题。
VIF的公式为：
个自变量与其余自变量不相关，表明不存在多重共线性。
一般而言：
VIF 越高，表明越有可能存在多重共线性，需要进一步研究。当 VIF 超过 10 时，存在需要纠正的显著多重共线性。
例如，假设一位经济学家希望测试失业率（自变量）与通货膨胀率（因变量）之间是否存在统计显著关系。如果将与失业率相关的其他自变量，例如新增初次失业救济申请，纳入模型，可能会引入多重共线性。
整体模型可能显示出强大的统计解释能力，但无法确定影响主要是由于失业率还是新增初次失业救济申请。这便是 VIF 能检测到的问题，它可能建议将其中一个变量从模型中剔除，或寻找某种方式合并它们，以捕获其联合效应，具体取决于研究者想要检验的假设。
作为经验法则，VIF 值在 3 以下不会引起担忧。随着 VIF 的增加，你的回归结果的可靠性将降低。
VIF 等于 1 意味着自变量不相关，回归模型中不存在多重共线性。
VIF 用于衡量回归分析中自变量之间的相关性强度。这种相关性被称为多重共线性，可能对回归模型造成问题。
虽然回归模型中适度的多重共线性是可以接受的，但较高的多重共线性则可能引发关注。
可以采取两种措施来纠正高多重共线性。首先，可以剔除一个或多个高度相关的变量，因为这些变量提供的信息是冗余的。第二种方法是使用主成分分析或偏最小二乘回归，而非普通最小二乘回归，这样可以分别将变量减少为无相关性的较小集合，或创建新的无相关变量。这将提升模型的可预测性。
[1] CFI. "Variance Inflation Factor."
[2] Isixsigma. "Variance Inflation Factor (VIF)."
TheQuantHackers是由一群来自世界顶尖高校和量化金融从业人员组成的前沿社区，致力于探索人工智能（AI）与量化（Quant）领域的无限可能。我们的团队成员来自剑桥大学、牛津大学、哈佛大学、苏黎世联邦理工学院、北京大学、中科大等世界知名高校，外部顾问来自Microsoft、HSBC、Citadel、Man Group、Citi、Jump Trading、国内顶尖私募等一流企业。
Upgrade to unlock all institutional-grade algorithms, derivative pricing engines, factor backtesting frameworks, and live QuantLab execution.
Explore Membership Access
Pre-order · £490Quantitative analytics platform
Waitlist closed — signal locked. Pre-order TQH TERMINAL now at £490 with launch-day delivery.