190 The Basics of financial economeTrics
■ (^) The appropriate number of lags can be selected either by using a partial
autocorrelation function or by using some information criterion. This
often requires considerable experimentation.
■ (^) A model is selected if the residuals of the model are white noise.
■ (^) It is important to evaluate models based on the accuracy of forecasts.
■ (^) A model’s forecast performance is judged adequate when its mean
squared error is small relative to competing models.
■ (^) Using vector autoregressions it is possible to model multiple time series
variables.