Non Linear Least Square fitting for large data

Dear all,

I would like to do least square fitting for a data I get from an antenna. The data looks like this Data plot. I tried using this function in Matlab to fit the data and I used a polynomial of 7 degree, but the fitting wasn't appropriate. People suggested that using a polynomial of a higher degree than 7 is not a good idea.

function leastCurvePolynomialFit(x,y, n)
p = polyfit(x, y, n)
poly_line = polyval(p, x);  
%Plot data and estimated polynomial
figure
hold all
plot(x,y,'+')
plot(x,poly_line)
xlabel('X')
ylabel('Y')
title('Noisy Data and Fitted Line')
box
end

If anyone can help me solve this problem without toolbox I will really appreciate it because I don't have any extra toolbox from Matlab and I would like to convert the code to C++ eventually.

Many thanks

[Star Strider]

Do you have a mathematical model of the radiation pattern of your antenna? If so, use the nlinfit and nlpredci (Statistics Toolbox) functions to fit your data to it and estimate the prediction errors. Use the prediction errors to eliminate outliers.

Answer 1

[dpb]

doc spline  % and friends

Knowing the end purpose of the fitted line may be of some help in refining a response.

[ADASI]

This data is in the form (X,Y) where X represent an angle from -180,+180 and Y represent the signal strength that I get from my antenna.I need to do a non linear least curve fitting to remove any outliers from the data.