Excel Vs Matlab Curve Fitting Tool

Hello to everyone whoever is reading this post,

I am using a curve fitting tool for bunch of data points and used 1 degree polynomial to test the equation and coefficients that which i also did in excel. The coefficients seem to be differ, I don't understand why, am i missing something here?

Result from the Excel :

Results from Matlab:

As you see the coefficient seems to be differ so much.

As for a tets example :

x variable 1 : 5

x Variable 2 : 10

Looking for = 0.28+5*0.041-10*0.0038 = 0.447 (From Excel)

Looking for = 0.7639+5*0.1439-10*0.001785 = 1.46 ( From Matlab) which is far off then Excel

Thanks

Karki

Answer 1

[dpb]

Read the output from Matlab -- the model has been standardized whereas the Excel model uses raw data. Note the R-sq and adjusted R-sq values are the same -- they both fit the same model identically.

[Steven Lord]

To be more explicit:

x = 5;
y = 10;
normalizedX = (x-13.75)./3.431;
normalizedY = (y-24.26)./4.631;
f = 0.7639 + 0.1439*normalizedX - 0.01785*normalizedY

That's in pretty close agreement to the value you received from Microsoft Excel. If you evaluate the fit with your x and y as shown in the "Evaluate the Fit at a Specific Point" section on this documentation page it will use the full double precision values of the normalization factors and the coefficients. The result of that direct evaluation of the fit should agree even more closely with Excel than the indirect evaluation using the rounded normalization factors and coefficients.

[Mahesh Karki]

Thank you Steven for your clear answer. Appreciated so much. :)