regress(x,y) for least square regression of two variables x,y

when i use regress(x,y),I obtain only one answer :shouldn't i get two answers which are the slope and y-intercept ?

[mohamed]

my code is x=[1 4 5 6];y=[4 6 7 9]

Answer 1

[the cyclist]

You don't show us your code, but I am guessing you neglected to add a column of ones to your x input, as described in the documentation:

doc regress

This code will give you the two-parameter output you expect:

x = [ones(3,1),rand(3,1)];
y = rand(3,1);
b = regress(y,x)

[mohamed]

my code is x=[1 4 5 6];y=[4 6 7 9] how can i use it with your code ?

[the cyclist]

x=[1 4 5 6];
y=[4 6 7 9];
b = regress(y',[ones(4,1) x'])

You need to add the vector of ones (for the constant term), and use column vectors rather than row vectors.

[mohamed]

i am very thankful to you but i wish to answer my last question : if i have another variable z=[ 30 40 25 7], how can i find the regression of z on y and x after combining them in one array ? i don't know how to combine them or on what basis ?

[Image Analyst]

But I did answer that here. Did you ever look at where you first asked it?

[the cyclist]

@Image Analyst, I answered there, too, and did not realize it was the same poster! sigh

[the cyclist]

b = regress(z',[ones(4,1) x' y']);

combines x and y into one array, adds the column of ones that you need, and then does the regression with z.

Answer 2

[Image Analyst]

I don't have regress(). Please list what Product (toolbox) is is in below your question in the Product box.

You can use polyfit() which is in base MATLAB to get the slope and intercept in one line. Here's a full blown demo:

clc;
% Create training data:
x = [1 4 5 6]
y = [4 6 7 9]

% Do the regression.
coefficients = polyfit(x, y, 1); % 1 means a line
slope = coefficients(1)
intercept = coefficients(2)
% We're done!

% Plot training data:
plot(x, y, 'bo', 'LineWidth', 3);
hold on;

% Plot fit
numberOfPoints = 100;
fitX = linspace(min(x), max(x), numberOfPoints);
fitY = polyval(coefficients, fitX);
plot(fitX, fitY, 'r-', 'LineWidth', 3);
grid on;
xlabel('X', 'FontSize', 30);
ylabel('Y', 'FontSize', 30);

[the cyclist]

I always forget that regress() is not in core MATLAB. I added the Statistics Toolbox product tag.