How can I build an trendline through a fixed point?

I have scatter data and need to fit it with a curve (line) which passes through one exact and fixed point (0,0).

How can I do this?

Thank's for respond.

Iaroslav.

Answer 1

[Star Strider]

If you do not include a constant (intercept) term in your model, it will force an intercept at (0,0). So if it is a straight line, to fit (x,y) data, this works:

x = randi(10, 10, 1);           % Create Data
y = randi(50, 10, 1);           % Create DAta

B = x(:)\y(:);                  % Regression Of Line Through Origin

yfit = x(:)*B;                  % Calculate Fitted Line

figure(1)
plot(x, y, 'bp')                % Plot Data
hold on
plot(x, yfit, '-r')             % Plot Fitted Regression Line
hold off
grid
axis([0  max(x)    0  max(y)])

[pfb]

this is definitely more professional :)

[Star Strider]

Thank you.

Answer 2

[pfb]

Is that a straight line or a curve? And, in the second case, which curve?

In the first case you're basically expecting that y is proportional to x. So y./x should be roughly constant, and mean(y./x) would be an estimate of the slope, right?

Answer 3

[Iaroslav Gritsayenko]

Thank you very much, I understood how to do it.

My line will have equation y= a*x+b, where b=0; And a=mean(Y./X)

And does mathlab have some function to build fitting curve (not only line) with the fixed point in general case?

[Star Strider]

MATLAB functions all want (or will insert) a non-zero intercept term in the regressions they calculate. If you want a general case that will always fit a zero intercept, the code in my Answer will provide it.

[pfb]

perhaps custom equation in cftool?