PESTS: Poisson Estimators for State-Space Time Series

Welcome to the PESTS homepage. PESTS was created by Patrick T. Brandt and John T. Williams.
PESTS is a series of GAUSS or R (GNU-S) programs for estimating event count time series models.


What can PESTS do?


Problems and Questions

Our work on event count time series models and other non-normal time series led us to realize that there are few readily available tools for social scientists who wish to investigate the time series properties of event count data. These programs will allow you to estimate the state-space time
series models discussed in our papers.

The software in PESTS is intended for to estimate two different time series models::a Poisson exponentially weighted moving average model (PEWMA) and a Poisson autoregressive model (PAR(p)).

PESTS is a series of GAUSS programs that work together as library or R source code. The GAUSS library is easy to install in your existing installation of GAUSS. The only requirement is that you also have a copy of MAXLIK,,the GAUSS add-on module for maximum likelihood estimation. If you wish to use the R version, you need only download the source file and examples.

You can have PESTS up and running in a matter of minutes.


We have developed both of the models implemented in PESTS in two papers:

Brandt, Patrick T., John T. Williams, Benjamin O. Fordham and Brian Pollins. 2000. "Dynamic Modelling For Persistent Event Count Time Series." American Journal of Political Science 44(4): 823-843.

Brandt, Patrick T. and John T. Williams. 2001. "A Linear Poisson Autoregressive Model: the Poisson AR(p) Model." Political Analysis 9(2): 164-184.

We also presented a short course on these methods at the 1999 American Political Science Association Meeting. You can see the video on an INTERNET2 by selecting the previous link.

You can download the slides from the course:

Brandt, Patrick T. and John T. Williams. 1999. Time Series Models for Event Count Data. Presented as a short course at the Annual Meeting of the American Political Science Association, Atlanta, Georgia. September 1, 1999.


PESTS allows you to estimate the two state-space time series models we have developed. These are the Poisson exponentially weighted moving average model (the PEWMA) and the Poisson autoregressive model (the PAR(p)). The PEWMA is intended for data that is persistent, while the PAR(p) follows a stationary AR(p) process. The Gauss programs we have developed require the the GAUSS add-on module MAXLIK, which performs maximum likelihood estimation. An example of data and diagnostics can be seen by selecting the following link:

A PEWMA example: Supreme Court Case Agendas


The PESTS files and the manual can be downloaded by clicking here.


The PESTS manual contains the following:


If you have any problems with your PESTS code, be sure to to read the section of the manual titled
"Troubleshooting and Problems." If you are still having problems, you can check our troubleshooting and problem page
to see if we have a possible solution. Otherwise, e-mail us with a detailed description of the problem. Please be
sure to include the information listed below:

  1. PUT "PESTS PROBLEM" in the SUBJECT LINE of your e-mail.
  2. LIST OF ERRORS: If Gauss is generating errors, please copy and send to us the exact error message
    that is being generated.
  3. WHICH MODEL(S) YOU ARE ESTIMATING: You might have the wrong specification, in which case
    the program can be very hard to use.
  4. WHAT YOU ARE USING TO OPTIMIZE THE LIKELIHOOD: This includes the method you are using
    to get starting values, the optimization algorithm, and any other MAXLIK options you have set.

All this information should be included in an e-mail addressed to PATRICK BRANDT. We will reply as soon as
humanly possible.