Variance of the OLS Estimator

Also known as · Variance Inflation Factor · Variance of Estimator

The sampling variance of β^\hat\beta. In the simple model it is σu2/∑(xi−xˉ)2\sigma_u^2/\sum(x_i-\bar x)^2; with several regressors it is σu2/[(1−Rj2)∑(xji−xˉj)2]\sigma_u^2/\big[(1-R_j^2)\sum(x_{ji}-\bar x_j)^2\big], where Rj2R_j^2 comes from regressing xjx_j on the other regressors.

When to use

Read it to see what makes an estimate imprecise: a large error variance, little spread in xjx_j, or strong Multicollinearity (Rj2→1R_j^2\to 1). The factor 1/(1−Rj2)1/(1-R_j^2) is the Variance Inflation Factor.

See Lecture 4 and Lecture 8.

Appears in