FORECAST ACCURACY AND ERROR EVALUATION

Two means of measuring forecast accuracy:

1. Examine ex post error terms

2. Simulate ex ante error terms

Graphical methods of model evaluation:

1. Time series examination of actual and potential values

2. Time series examination of error patterns

3. Control charts for cumulative error terms (a measure of cumulative bias)

4. Plot of actual versus predicted values (line of perfect forecasts)

5. Diagram of turning actual versus predicted change (turning point accuracy)

Statistical methods of model evaluation

1. Proportion of false signals and missed signals

Actual	No TP	TP
No TP	NN (correct)	NT (wrong)
TP	TN..(wrong)	TT (correct)

Proportion of false signals = NT/(NT + TT)

Proportion of missed signals = TN/(TN + TT)

2. Mean squared error and root mean squared error

3. Mean absolute error

4. Mean absolute percentage error

5. Mean error (a measure of bias) and mean percent error (a measure of bias)

6. Root percent mean square.