Time series decomposition

Question

Hello,

a customer told me that in an older version of JMP there was a functionality to decompose time series. The data has one time column and one measurement column. The analysis then tells how many significant seasonal trends could be overlaid to fit the data and what the respective periodicity is. E.g. the curve may be build by adding a 3-hour season, a 7 hour season and a 10 hour season and the rest is white noise. He couldn't remember to have specified a model or other parameters.

I couldn't find any timeseries functions that deliver something like that nor can I imagine other platforms that do something like this.Any hints?

Thanks and best regards

Bernd

jiancao · Answer

Mia is right that time series decomposition (aka X11) is a new feature in JMP 12. But I think your customer might refer to Spectral Density option in JMP's Time Series platform.

There is also a recently added JMP Addins for Frequency Analysis and Dual Seasonality Time Series Analysis

It enhances the spectral density analysis in JMP today.

mia_stephens · Answer

Hi Bernd,

Time series decomposition is a brand new feature in JMP 12.  We (academics) have been requesting this for some time - to my knowledge this has never been directly available in JMP.  Maybe the customer was using an add-in or a script (although, I'm not aware of any publicly available)?

Hope you're well!

Mia