This message has been cross posted to the following eGroups: Business and Economic Statistics Section and Statistical Consulting Section .
-------------------------------------------
I have two estimation problems that I need help with. Please my inarticulateness in advance.
The first problem is a time series that is essentially flat, i.e., zero-sloped, up to a certain period, then takes off with a positive slope for all subsequent periods. I think it would be optimal to estmate the flat section and the sloped section of the time series simutaneously, but I would like to do so under two different approaches. The first approach would take whatever appears as the break point as a given and estimates the two 'sections' of the time series simultaneously. The second approach would simultaneously estimate the flat and sloped sections, but would also simultaneously optimize over the break point. What estimation technique(s) would I use for each of these two approaches?
Here's the second problem. I have a set of time series that are essentially all a function of a single variable. The set is clearly both contemporaneously correlated across series and each one is serially correlated. I would like to forecast the set of time series based on projections of the single variable that each of the series is dependent upon. I've thought about vector autoregression, but that seems not to capture the essence of the problem correctly. Any suggestions?
Any and all help is greatly appreciated.
Phil
-------------------------------------------
Philip Hanser
-------------------------------------------