Hi,
a customer of mine has the following goal:
Based on different predictors (health status, habits, ...) they want to build a model to predict desease/death. The data base is 10y aggregated data.
Now they add economic factors like GIDB, Country, population ... and want to forecast the timeseries. That is looking from now into the future for 1 or 2 years and see how the desease rate/morbility rate changes over time by changing the predictors and additional factors. So I#m not just interested in the probability of death for a specific setting but the change over time.
I wonder if we have already some example data I could use to play with the data to understand the different options of methods and models in advance to talking to the customer.
From an econometric point of view I probably would use a kind of autoregressive or VARMAX model (but this would be more challenging to do in JMP)
Thanks for suggestions (and/or data, journals, ...),
Martin
Martin,
You are correct--that'd be better handled by autoregressive models. But you might want to give Transfer Function Models a try http://www.jmp.com/support/help/13-2/Statistical_Details_for_Transfer_Functions.shtml#428769
There is a sample data set, SeriesJ, in JMP, which was used initially by Box and Jenkins. (as a cautionary note--I have found it difficult building a Transfer Functions model with multiple input series like you described.)
Jian
