Linear Regression with Only Dependent Variables

Hi,

So I have 2 matrices that are both nx7. Each column is associated with a tone that is played in both matrices. I want to compare the first columns from both matrices in a scatter plot and do (I think) a linear regression to see how they are correlated. Based on my understanding of the fitlm function, I need a matrix of observations and a column vector of predictors, and I don't have that, I have 2 column vectors instead. My question is how do I properly run the linear regression, if that is what I am supposed to be running. Thank you!

Answer 1

[Jeff Miller]

It sounds like you can get what you want from simply

[rho, pval] = corr(matrixA(:,1),matrixB(:,1))

fitlm is for situations with a variable of primary interest that is to be predicted and other variable(s) that are used as predictors. Your variables sound more symmetric/equivalent than that.

Answer 2

[Torsten]

The model with parameters a, b and c to be determined should be

column_matrix_1 * a/sqrt(a^2+b^2) + column_matrix_2 * b/sqrt(a^2+b^2) + c == 0

You can use "lsqnonlin" to fit the coefficients.