#coefficient-variance

Question 1: Single choice

mediumAI-checkedLinear Models & Regularization

In an OLS model with features $x$ , $y$ and $z$ , the column $z$ equals $x + y$ plus a little random noise, so the three features are almost collinear. What is the typical consequence?

The predictions on new data from the same distribution become much less accurate1
The fit fails with an error because $X^\top X$ cannot be inverted2
The three coefficients get large standard errors and vary a lot between samples3
The coefficient of $z$ is automatically forced to exactly zero4

#coefficient-variance

Timed test