Algorithms for functions (LINEST, RSQ and LOGEST)

[Guest]

I am looking for the algorithms underlying Excel 2003's LINEST, RSQ and
LOGEST functions. Are these algorithms published somewhere? If so, where?

 

[Myrna Larson]

Search the MS Knowledge Base. If they are published, you should find them
there. www.support.microsoft.com is the place to start.

 

[hgrove]

Myrna Larson wrote...

Search the MS Knowledge Base. If they are published, you should fin them
there. www.support.microsoft.com is the place to start.

The best hit for LINEST and Excel 2003 is

http://support.microsoft.com/default.aspx?scid=kb;en-us;828533#kb1

but I wouldn't say the explanation is simple to follow

 

[Jerry W. Lewis]

The algorithm for RSQ is documented in Help; it uses CORREL()^2.

LOGEST(ydata,xdata) = LINEST(yData,LN(xData))

LINEST traditionally used the matrix equation for estimates of
(X'X)^-1 X'y
The problem was that if X is ill-conditioned, then X'X is even more
ill-conditioned (possibly even numerically singular). (This same issue
is more easily seen with the 2003 improvements to VAR, etc.) 2003 uses
the same underlying concept, but decomposes X to permit analytical
cancellation and numeric inversion of a less difficult matrix. The
result is potentially far more accurate than in previous versions.

There are some reported anomalies, suggesting that LINEST/LOGEST
coefficient estimates that are exactly zero in 2003 should be looked on
with some suspicion.
http://groups.google.com/groups?threadm=412980D4.5040305@no_e-mail.com

Jerry