Problem using polyfit in a loop

Hello

I am having problem using polyfit in a for loop.

A is a null matrix. Polynomial fitting of order 10 has to be subtracted from the raw data z((I(:,j)+100/dt:end),j) and the residual has to be zero padded. t(I(:,j)+100/dt:end) is the time axis. But the loop shows matrix dimensions do not agree. Please find the code below.

for j = 1 : 1 : L
  z1(:,j) = A(:,j) + (padarray(z((I(:,j)+100/dt:end),j)-polyval(polyfit(t(I(:,j)+100/dt:end),z((I(:,j)+100/dt:end),j),10),t(I(:,j+100/dt:end)),j)),N-length(z((I(:,j)+100/dt:end),j),0,'post');
end

[dpb]

You don't have to try to write everything in one line...factor your code first so you can read it and debug...once have done that, use the debugger to step through and see where your logic error is.

[shaina dhamija]

Thank you for your response. I could debug it by not writing everything in a line. Found the error.

Answer 1

[Steven Lord]

If you're trying to remove a polynomial trend from your data, see the detrend function. As of release R2019a you can specify a degree >1 of the polynomial trend to remove.

[shaina dhamija]

Thank you for responding. I wil check the detrend function if it works for me.

[dpb]

High-order polynomials are notroiously ill-behaved and prone to numeric issues...depending upon what range of values are for x, x.^10 can become extremely large. Consider using some other form than such a high-order polynomial or at least use the option in polyfit|polyval to standarize the inputs that will help on the numerics...