Department of Statistics Seminar
North Carolina State University
presents
Dr. David F. Findley
Bureau of the Census
Title: "Model Selection via Out-of-Sample Forecast Error Comparisons: Practice and Theory"
ABSTRACT
Many monthly economic time series can be effectively modeled with autoregressive moving average models that include regressors for the effects of moving holidays, weekly activity patterns, and outliers. Many of the model comparisons needed by the modeler of such data are non-nested comparisons for which there are no practical hypothesis tests. We first show a few empirical results demonstrating the versatility of a graphical model selection diagnostic available in the Census Bureau's X-12-ARIMA and X-12-Graph software that compares the relative abilities of two models to forecast withheld data.Then we present a theoretical result on the a.s. convergence of average squared forecast errors from recursively estimated models with incorrect ARMA and/or regression components. This result suggests some of the diagnostic's observed behavior. Both OLS and iterative GLS estimates of the regression coefficients are covered.
Friday, February, 16, 2001
3:35 - 4:35 pm
206 Cox Hall
Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.