Graphing Autocorrelations

Autocorrelations and partial autocorrelations are bounded between -1.0 and 1.0. Graphing them without fixing the graph range can produce confusion, because the plot ranges for different sets of correlations will probably be quite different. We have found the following set of options to be very helpful:

graph(number=0,style=bargraph,max=1.0,min=-1.0)

# corrs

NUMBER=0 causes the time axis to be labeled with 0,1,2,etc. STYLE=POLYGONAL also works well for correlations, and STYLE=SPIKE can be useful if you have to graph a large number of correlations. See the section on Box-Jenkins models for more details on computing and graphing correlations.

This shows a very basic graph of correlations. If you use any RATS procedures which compute and graph autocorrelations (such as the @BJIDENT procedure), you are likely to run into some graphs that supply additional information. For instance, it’s possible to highlight the correlations which are statistically significant using the SHADING option.