Complex values with nlinfit

Hello everyone,

my code looks like this:

%initializing values
u_star = .2; 
kappa = .4;
z_0 = 10^-6;
d = 0;
alpha = .11;
z = [33 40 50 60 70 80 90];

%sector means
m_smmr = [6.5261 6.6065 6.9178 6.9645 7.1122 7.3116 7.4411];
m_wntr = [10.5684 10.7004 11.2594 11.4037 11.6157 11.9577 12.1230];

%modelfunction
wpfun = @(a,z) (a(1) / kappa)*log((z-d)/a(2));

a = [u_star z_0];

wpsmmrfit = nlinfit(z,m_smmr,wpfun,a);
wpwntrfit = nlinfit(z,m_wntr,wpfun,a);

I expect real values for both fits, but the wpwntrfit is giving me a complex output. My outputs look like this:

wpsmmrfit =

    0.3645    0.0269


wpwntrfit =

   0.6349 - 0.0000i   0.0441 - 0.0000i

I have no idea why or whats going on. How do I make the wpwntrfit have non complex outputs?

Kind regards,

Kai

[Sulaymon Eshkabilov]

If your chosen model logarithmic function is correct or what you need to use, then the data (Y) m_wntr need to be adjusted. By changing the values of m_wntr your fit model coefficients will be real. E.g.:

wpwntrfit = nlinfit(z,m_wntr-5,wpfun,a)

0.6349 1.0295

Or you'd need to change the model function type: wpfun

[Kai Hillers]

Thank you for your answer!

Can you elaborate why the values of m_wntr are too big for the modelfunction? How would i have to change the modelfunction if I'm not changing the data.