Autocorrelation can arise from a
number of different factors. In time
series data, autocorrelation is often attributed to normal business
cycles. However, the absence of
significant variables in the regression equation or nonlinear relationships can
also produce this undesired effect.
There are two types of Autocorrelation. Figure (a) exemplifies positive
autocorrelation while figure (b) shows an example of negative
autocorrelation. Notice that in the
case of positive autocorrelation, successive disturbance terms (this is the
residual error plotted against time) tend to be followed by disturbances of the
same sign. Just the opposite is the
case for negative autocorrelation; successive error terms have opposite signs.
