Linear weighted regression

[oliver]

Hi all,

I would like to perform a weighted linear regression. Does
anybody know about an add-in that is able to perform such
an analysis?

Unfortunately, the Data-Analysis-Add-In is only good for
unweighted regressions.

Thanks a lot for your help!!

oliver

[Jerry W. Lewis]

Calculate a weighted average by

=SUMPRODUCT(<y-data>,<weights>)/SUM(<weights>)

(handle x-data analogously). The slope of weighted linear regression is
then

=SUMPRODUCT(<y-data>-<y-wtd_ave>,<x-data>-<x-wtd_ave>,<weights>) /
SUMPRODUCT((<x-data>-<x-wtd_ave>)^2,<weights>)

and the intercept of weighted linear regression is

=<y-wtd_ave> - <x-wtd_ave>*<wtd_slope>

These should work if there is no missing data. With missing data, you
must insure that the averages do not include unmatched data.

Jerry

[boxa...@g.cofc.edu]

Hi Jerry,

If I want to do a linear regression with 1/x weighting, how do I calculate the weights for the X data and Y data?

Thanks,

Baylye

[Auric__]

You start by not replying to a post that's ***THIRTEEN*** FUCKING YEARS OLD!

--
- I'm not brave. I'm just too tired and hungry and angry to care anymore.
- Perhaps they are not mutually exclusive. Perhaps bravery is simply apathy
with delusions of grandeur.

[Jambo Frigerio]

Dear Auric,

I am very sorry for your unfriendly answer to Baylye.

Weighted linear regression is still a widely used approach in analytical chemistry and maybe in other fields.

No matter if the post is 13, or 50, or 100 years old!

Baylye's question was very interesting and an answer to it would benefit other interested people.

Best regards,
Gianfranco