From: matthew@spielman.com
Date: 11 Jun 2003
Time: 22:50:18
I'm working on documenting the statistical methods used by an old software package made by Deloitte & Touche in 1995 for estimation sampling of inventory.
The program spits out, among other things, a point estimate of the inventory level accompanied by a confidence interval which we usually run at 95%.
The program says it runs a "combined regression" which is not something that I am familiar with but the point estimate formula is the same as with a multiple regression.
The problem is that the documentation says that it computes the confidence interval using: y-hat (+/-) t x (Standard Error of the Estimate) however, I can not figure out what formula they are using for the Standard Error of the estimate. Using the formula that I am familiar with (square root of (SSE/(n-2)) I get a figure which is SIGNFICANTLY lower than what the program produces using the same data.
I am assuming that I am just using the wrong formula but i definitely need to find out because on the odd chance that I am right, we producing far too broad an interval for a 95% confidence level.
Thanks, Matt