Subscribe Bookmark RSS Feed

Issue with Simple ARIMA model - multivariate

stephendonnelly

Community Trekker

Joined:

Sep 19, 2013

Hi everybody,

I'm very new to JMP and have had some success with basic tasks but am struggling with AR modelling.

Specifically I'm hoping to perform a H(heterogenous)AR model but I'm not sure the capability exists.

Long story short, I'm building a realised volatility model of oil price. The literature suggests you create the model something along the lines of: RV = a + bRV(5) + bRV(10) + bRV(22) + e. The different betas represent realised volatilities over different time horizons, i.e. 5 days.

I build the time series ARIMA (1,1,1) model with RV as the y and the other three RV's as the x's. However the output I get is 4 individual models, rather than one complete model with all my parameters.

I'm trying to replicate the process that my lecturer did within R but so far I'm struggling and 4 individual outputs doesn't provide me with anything useful. I've just started my new job and they would like me to produce this model.

Do you have any ideas?

If you need any further information I'm more than happy to supply it.

Thanks in advance,

Stephen

P.S.   Is there anyway to increase the amount of observations JMP will process? My data is 5 minutely and overwhelms JMP. I've just sliced off a sample to perform the modelling in the mean time.

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

Hi Stephen,

Can you provide some more details about what you've done and what output you got?

Are you using the Time Series platform? When you say that you used "the other three RV's as the x's," which role in the dialog did you put them in?

4206_JMPScreenSnapz002.png

If you put them in the Input List in the Time Series launch dialog you will initially see each of the variables (Y and Input List) in the report. This gives you the opportunity examine each column individually and determine the best time series model for each series.

4207_JMPScreenSnapz003.png

Then you can use that information when you choose Transfer function from the red triangle menu at the top of the report.

4208_JMPScreenSnapz004.png

This will give you a dialog where you can specify the transfer function model.

4209_JMPScreenSnapz006.png

Here's an example of what the output from a Transfer Function Model looks like.

4210_JMPScreenSnapz007.png

I hope that helps.

-Jeff

-Jeff
2 REPLIES
Solution

Hi Stephen,

Can you provide some more details about what you've done and what output you got?

Are you using the Time Series platform? When you say that you used "the other three RV's as the x's," which role in the dialog did you put them in?

4206_JMPScreenSnapz002.png

If you put them in the Input List in the Time Series launch dialog you will initially see each of the variables (Y and Input List) in the report. This gives you the opportunity examine each column individually and determine the best time series model for each series.

4207_JMPScreenSnapz003.png

Then you can use that information when you choose Transfer function from the red triangle menu at the top of the report.

4208_JMPScreenSnapz004.png

This will give you a dialog where you can specify the transfer function model.

4209_JMPScreenSnapz006.png

Here's an example of what the output from a Transfer Function Model looks like.

4210_JMPScreenSnapz007.png

I hope that helps.

-Jeff

-Jeff
stephendonnelly

Community Trekker

Joined:

Sep 19, 2013

Hi Jeff,

Sorry if I didn't provide the detail correctly, you interpreted what I meant to say correctly though.

Simply put you answered my question exactly, I was getting output just like your second image but didn't know how to get it to the fifth image.

Thanks for your detailed answer with screenshots included. It was a big help.

Best,

Stephen