Polyfit and polyval in Matlab

Hi!I have a problem with polyfit and polyval. I have two vectors (x and y). You can see it in the attached image (the red line). I want to create a polynom that allow to create the blue line (imagine that it doesn't exist) only with the values that I have.

http://imageshack.us/photo/my-images/6/graphrw.jpg/

I've tried with polyfit and polyfit but it doesn't work at all.

x2=x'; %I have changed the direction because if not, it doesn't do the polyfit.
coeffs1=polyfit(x2, msensiv_graf,1);
coeffs3=polyfit(x2, msensiv_graf,3);
coeffs5=polyfit(x2, msensiv_graf,5);
coeffs12=polyfit(x2, msensiv_graf,12);
new_y1 = polyval(coeffs1, x2);
new_y3 = polyval(coeffs3, x2);
new_y5 = polyval(coeffs5, x2);
new_y12 = polyval(coeffs12, x2);

   plot(x2, msensiv_graf, 'r');
   hold on
   plot(x2, new_y1, 'r');
   plot(x2, new_y3, 'b');
   plot(x2, new_y5, 'g'); 
   plot(x2, new_y12, 'y');

And the error is:

Warning: Polynomial is badly conditioned. Add points with distinct X values, reduce the degree of the polynomial, or try centering and scaling as described in HELP POLYFIT.

Does somebody know what can I do? I also think to try interpolation to get the blue line.

Thanks a lot in advance.

Answer 1

[Image Analyst]

Try using the S and mu outputs of polyfit(), like it (almost) says. I once had the same error and that helped me. Just get them and pass them into polyval() - it was as simple as that.

[Image Analyst]

So, Emmanuelle, did that work for you?

[Emmanuelle]

This is my result. The red crosses are my points and the lines, is what I obtain with the following code:

coeffs1 = polyfit(x, nuevo, 1);
coeffs100 = polyfit(x, nuevo, 100);
 new_y1 = polyval(coeffs1, x);
 new_y3 = polyval(coeffs3, x);
 new_y100 = polyval(coeffs100, x);
 plot(x, nuevo, 'rx');
 hold on
 plot(x, new_y1, 'k');
 plot(x, new_y100, 'g');

[Image Analyst]

Can you supply the numbers for x and y so we can run it? By the way, a polynomial of 100 is not much of a regression is it? Plus, I don't see that you even have 100 points so how can you do a 100th order "fit"? And where did coeffs3 come from? It seems some code is missing.

[Emmanuelle]

Sorry, the code is the following but I don't get any good result (I need a line that cross the different points, like an interpolation). Am I doing something wrong?

[coeffs1,S,mu] = polyfit(x, nuevo, 1); new_y1 = polyval(coeffs1, x); plot(x, nuevo, 'rx'); hold on plot(x, new_y1, 'k');

[Image Analyst]

Did you read my prior comment about supplying input data we can work with?

[Emmanuelle]

Image Analyst, thanks for your reply. Yes,I've changed but it doesn't work at all. Now I'm trying directly with interpolation... (this is a capture of my result)

And this is what I want to get, like the example in Matlab's web:

My code:

• plot(x2, nuevo); xt= 250:0.1:2500; yt= interp1(x2, nuevo, xt, 'spline'); plot(x2, nuevo, 'o', xt, yt);*