Curve Fitting

I've been trying to fit different series/methods to a certain set of data. I have a polynomial fitted to the data: P = polyfit(x,y,3)

but I cannot get the fourier or the gaussian commands to work: p = fit(x,y,'fourier3)

Can someone help me?

Thanks

Answer 1

[Teja Muppirala]

As proecsm points out, your data needs to be made into columns. Like this:

x = 1:1000; %Row vector
y = sin(0.01*x) + cos(0.03*x) + 0.1*randn(size(x)); %Row vector
F = fit(x(:),y(:),'fourier3') %Use (:) to turn it into a column
plot(x,y)
hold on
plot(F)

Also, are you aware of the Curve Fitting Tool GUI?

cftool

This makes it easier to try out different fits interactively (for both curves and surfaces), and you can automatically generate MATLAB code from your results.

Answer 2

[Brandon Armstrong]

Does not work mean you get an error, or just that the fit is very bad? If it is not fitting properly you can try giving some initial guesses and and upper and lower bounds for your coefficients using 'fitoptions'.

Answer 3

[Nathan]

I'm getting these errors whenever I try to use other fits besides polyfit

Error using fit>iFit (line 129) X must be a matrix with one or two columns.

Error in fit (line 109) [fitobj, goodness, output, convmsg] = iFit( xdatain, ydatain, fittypeobj, ...

Error in TestProcedure_2 (line 34) [p] = fit(x,y,'fourier4');

[Ali]

correct form: fit(x(:),y(:),'fourier4')

(:) are used to turn it into a column

Answer 4

[bym]

polyfit()

will work with both row-oriented or column-oriented data, whereas the documentation specifically mentions columns for

 fit()

I don't have the curve fitting toolbox to test this.