Outlier Diagnostics for Eviews 5.1

Below are links to two files for Eviews 5.1 (a program ".prg" and workfile ".wf1"). The program computes a number of diagnostics for outlier analysis, including h, rstudent, rstandard, dfits, covratio, Cook's d, Mahalanobis d, Welsch's d, and DFBETAs.

If you run the outlier_ford.prg program with the outlier_ford.wf1 workfile, you will see how it works. The end result is a group of series (outliers) that contains all the diagnostic measures. DFBETA is computed for each X, and will be labelled DFBETA_Xi for all i series.

CAUTION: This program will not work if there are missing values, but you can remedy that by moving all good observations onto a new sheet in Eviews 5.1. Also, make sure that the program is not overwriting or deleting any of your series, vectors, matrices, etc.

For the data included in the workfile, the output created by the program matches the output of STATA 8.0 (and the data in wf1 is from the auto.dta example file from STATA). The Excel spreadsheet "outlier_ford.xls" below shows the comparison (MahalD and DFBETA for the constant are not computed by STATA). I can't guarantee this program will work with all datasets. If you do find a problem, have some better code, or have any thoughts, drop me a note (gford@aestudies.com). You can get more information on Eviews at www.eviews.com.

outlier_ford.prg

outlier_ford.wf1

outlier_ford.xls