It is OLS in all versions. Prior to 2003, It formed the normal
equations and then inverted X'X ... Numerically, if you form X'X, you
have already lost the battle with ill-conditioned problems (finite
precision problems associated with squaring/summing large numbers).
Excel 2003 decomposes X, so that the solution to the normal equations
can be calculated without ever calculating the normal equations themselves.
Jerry
Jason said:
for Excel 2000, is it just a simple OLS regression? Is
this also true for 2000 add-ins, or do newer add-ins for
2000 use newer algorithms?
Jason
-----Original Message-----
...
Tools
Before Excel 2003, it comes straight out of standard
stats textbooks (which was
a problem). Beginning with Excel 2003, it uses a better
numerical algorithm, but
this means a single formula would no longer truly
describe it.