13 Regression models with autocorrelated errors
The standard assumptions say that the errors are uncorrelated. When the data are collected over time, such as monthly sales, yearly rainfall or daily prices, that assumption often fails. A shock that pushes this month’s response above the regression line tends to persist into next month, so neighbouring errors are correlated. The errors are then said to be autocorrelated (or serially correlated). This section shows what autocorrelation does to least squares, how to detect it, and how to fit and forecast when it is present.
13.1 Sources and effects of autocorrelation
13.2 Detecting autocorrelation
13.3 Estimating the parameters
13.4 Forecasting
13.5 Practice problems
13.2 Detecting autocorrelation
13.3 Estimating the parameters
13.4 Forecasting
13.5 Practice problems
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.