*To*: mathgroup@smc.vnet.net*Subject*: [mg11939] incorrect results*From*: "H.J. Wang" <hungjen@umich.edu>*Date*: Thu, 9 Apr 1998 00:33:47 -0400*Organization*: University of Michigan

Hi, I used David Belsely's Econometrics.m package --available from MathSource as well as in the book "Economic and Financial Modeling with Mathematica"-- to do some very simple Ordinary Least Square (OLS) regressions. However, using a small data set I found that the returned answer is very different from other statistical packages such as Stata or SAS. I then wrote my own OLS program in Mathematica, which is simply (X'X)^(-1) (X'y) where X are Nxk independent variable matrix and y is a Nx1 vector. The result is the same as that from Belsely's program, which means it is very different from other packages' results. I am wondering how this would happen. Does it have something to do with the working precision? If so, how could I make the adjustment? I am using Mathematica 3.0 in Win95. Thank you in advance for any advice! H.J. Wang