Polyfit and polyval plot intercepting zero

Hi Folks

I have a question about the polynomial regression function polyfit. Lets say that we have two random vector, for example; a = [2.1484, 2.1118, 2.0759, 2.0550, 2.0106]; b = [20.530, 20.221, 19.972, 19.721, 19.693];

If we want to do a polynomial regression it is very easily done by the inbuilt function: p = polyfit(a,b,n) If I want to plot the points and the polynomial curve it is also very easily done by the polyval function;

p = polyfit(p,a,n); q = polyval(p,q); plot(a,b,'r',a,q,'b')

(I think this is the right code, this first part illustrates the points in read and the second part the polyval function in blue) But I FOUND IT VERY HARD to make an polynomial regression which is forced to intercept the zero (this is in contrary pretty easy in excel). How can I achieve the polynomial evaluation, polyval, for an curve that intercept zero and easily put the function in a plot command with the points from vector a and b above? By the way, if it is possible, would it be possible to also combine all this, dots from a and b vectors, the polynomial through zero in a plot for standard deviation (the errorbar(a,b) function) and plot standard deviations for all the points from a and b.

Would really appreciate answer, have a nice day, regards from Sweden

Answer 1

[Torsten]

http://www.mathworks.com/matlabcentral/fileexchange/35401-polyfitzero

Best wishes

Torsten.

[TheLast One]

Thanks!

[TEDDY OKOLO]

Thanks! This was extremely helpful.

Answer 2

[math man]

Another, more vulgar way to solve this is to simply force where you want your Origin to be (Y at 0):

Where your varargin is a set of forced input pairs such as [0,0].

function [XY,Fit] = XY_ForceAndFitPoly(XY,PolyParamCount,varargin)

ForcedPairsRow = cell2mat(varargin);

ForcedPairsCount = size(ForcedPairsRow,2)/2;

for pair = 1:ForcedPairsCount

ForcedPair = ForcedPairsRow(1,1+(pair-1)*2:2+(pair-1)*2);

ForcedBlock = repmat(ForcedPair,10000,1);

XY = vertcat(XY,ForcedBlock);

end

Fit = polyfit(XY(:,1),XY(:,2),PolyParamCount);

XPlotVals = linspace(-.2,.2,31);

FitVals = polyval(Fit,XPlotVals);

hold on;

plot(XPlotVals,FitVals,'bo')

end