Excel/Linest fail

Advertisement


Question

Linest Fail
I am using linest function to compute coefficients of a 4th order poly to fit a data set. I have used it successfully but today I came across a data set which produces faulty coefficients. When I checked it with a graph and a trend line of 4th order poly, it gives the coefficients correctly. What could go wrong?"


Answer
Hi Kaushalya,

I expect this is due to the nature of your data. The Y values are all within a close range with relatively little variation and the order of magnitude of X and X varies considerably. This makes the computation of the least squares error-prone.

You will probably benefit from normalising your data, e.g. by subtracting the average of x from the x's and of y from the Y's. Then do the 4th order regression using the transformed data.

Of course you'll have to do some math afterwards to get the true formula (and coefficients) back.
About Excel
This topic answers questions related to Microsoft Excel spreadsheet (or workbook) stand-alone or Mircrosoft Office Excel including Excel 2003, Excel 2007, Office 2000, and Office XP. You can get Excel help on Excel formulas(or functions), Excell macros, charting in Excel, advanced features, and the general use of Excel. This does not provide a general Excel tutorial nor the basics of using a spreadsheet. It provides specific answers to using Microsoft Excel only. If you do not see your Excel question answered in this area then please ask an Excel question here

Jan Karel Pieterse

Expertise

Excel and Excel/VBA questions

Experience

Excel MVP

Organizations
Self employed Excel developer

Education/Credentials
Bachelor in Chemical Engineering

Awards and Honors
Microsoft MVP award since 2002

Past/Present Clients
Shell, Fortis bank, ABN-AMRO bank, Morgan Stanley, ...

©2012 About.com, a part of The New York Times Company. All rights reserved.