Generating White Noise and Stationary Plots

Peter_Polito

In SAS, there is a function to generate white noise and stationary plots when analyzing time series (https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/etsug/etsug_tfstart_sect028.htm), which show (graphically) probabilities of the Ljung-Box chi square statistic. In JMP, the Ljung-Box Q statistic is visible be default but there doesn't seem to be a way to get the corresponding graphs that SAS can generate. Is there a workaround to generate these plots and/or is there a scriptable solution? 

Thanks.

 

peng_liu

Do you mean that you want the bar chart representation of the Ljung-Box Q statistics in the Residuals report of a time series model?

I don't see anyway other than scripting. But I don't know how to write from scratch, but here is what I can do to piece things together:

1. First you need to get your time series modeling done.
2. Right click on the Ljung-Box statistics, and choose "Make into Data Table". The generated data table has a script, which tells me how to generate the table, along with how we get the analysis. (I am surprised by this buy-one-get-one-free.) We have the first piece.
3. Now use Graph Builder to draw the bar chart as you want, and produce a script after you get the chart that you want. We have the third piece.

For example, here is the GB script

Graph Builder(
	Variables(
		X( :"Ljung-Box Q"n ),
		Y( Transform Column( "Row", Ordinal, Formula( Row() ) ) )
	),
	Elements( Bar( X, Y, Legend( 6 ) ) ),
	SendToReport(
		Dispatch( {}, "Row", ScaleBox, {Reversed Scale} ),
		Dispatch( {}, "Y title", TextEditBox, {Set Text( "Lag" )} )
	)
);

that generates this chart

Then I stitch codes together. And here what I get

dt = open("$sample_data/time series/seriesg.jmp");

Names Default To Here( 1 );
platform = Data Table( "Seriesg" ) << Time Series(
	Y( :Log Passengers ),
	Variogram( 1 ),
	Seasonal ARIMA( 0, 1, 1, 0, 1, 1, 12, No Intercept( 1 ) )
);
Wait( 0 );
dt2 = 
Report( platform )[Outline Box( "Time Series Log Passengers" )][
Outline Box( "Model: Seasonal ARIMA(0, 1, 1)(0, 1, 1)12 No Intercept " )][
Outline Box( "Residuals" )][Table Box( 2 )] << Make Into Data Table;
Report( platform ) << Close Window;

dt2 << 
Graph Builder(
	Variables(
		X( :"Ljung-Box Q"n ),
		Y( Transform Column( "Row", Ordinal, Formula( Row() ) ) )
	),
	Elements( Bar( X, Y, Legend( 6 ) ) ),
	SendToReport(
		Dispatch( {}, "Row", ScaleBox, {Reversed Scale} ),
		Dispatch( {}, "graph title", TextEditBox, {Set Text( "Ljung-Box Q" )} ),
		Dispatch( {}, "Y title", TextEditBox, {Set Text( "Lag" )} )
	)
);

I cannot put line numbers above, but I can tell you that I added line 1, 10, and 16. All other lines are direct pasted into. Seeing in this screenshot.

At the moment, Time Series platform has limited support of unit root tests, and does not produce for residuals automatically. If you want it, you need to save residuals first, then run Time Series just on residuals. Several unit root test statistics are on the top right of the report. There are not p-values go with them either.