Calibration with lsqnonlin

Hi

I need to calibrate some parameters by minimizing the following equation: AP= sqrt(sum from i=1 to n w_i (P_i model - P_i market)^2) where w_i is a weight for each observation. I use lsqnonlin. I works fine without the square root, since I run a loop for each observation, and I have less parameters than observations.

The problem comes, when I sum up after the loop since lsqnonlin then think that I only have one observation and several parameters. Does anyone knows, how I can overcome this problem?

Thanks!

[Alexander Schneider]

I know, that the question was asked almost 8 years ago, but apparently people are still searching for topics like these.

The problem in the case above was, that the author returned a scalar instead of the full vector. Hence Matlab thought that the calibration problem was underdetermined and returned an error.