# Independence of Error Term Observations

Learn about independence of error term observations.

## We'll cover the following

## Correlation in observations

The Gauss-Markov theorem assumes that the observations of the error term are independent and uncorrelated. However, observations that are collected over time or across regions are likely to be correlated over time or within a region. If the temporal or spatial dependence isn’t modeled explicitly, the residuals will be correlated over time or spatially, causing standard error estimates to be biased. Since our focus in this course is the cross-sectional design, we won’t explore the time dependence, usually referred to as autocorrelation in the residuals. The dataset `final.85v2`

does have a spatial dimension. So, we’ll explore whether there’s any systematic pattern in the residuals that are correlated with regions. The figure below plots residuals versus fitted values by `continent`

, a variable for regions.

Get hands-on with 1200+ tech skills courses.