A discrete valued time series model;
The polio dataset

ADMB Files
Code: polio.tpl
Data: polio.dat
Initial values: polio.pin
All required files (DOS): polio.zip
All required files (linux): polio.tar.gz
Results: polio.par

Running ADMB-executables
In a DOS window
Under linux

Command line options:
-ilmn 5
-l1 10000000 -l2 100000000
-l3 10000000 -nl1 10000000

Results: Computation times
ADMB-RE: 66 seconds.
Kuk & Cheng: 3120 seconds.

ADMB-RE home
Otter Research

Model description

As an example of a discrete valued time series we use the 'polio data' considered by Kuk & Cheng (1999). It is assumed that yi has a Poisson (li) distribution, where

log(li) = Xib + ui.

Here, Xi is a covariate vector, b is a vector of regression parameters and ui is a first order autoregressive process

ui = r*ui-1 + ei.

Further details about the model can be found here: polio.pdf.