## nlinfit() not evaluating my power-law function

### Oliver Andersen (view profile)

on 18 May 2018
Latest activity Commented on by Oliver Andersen

on 19 May 2018

### Torsten (view profile)

Hello everybody!
I have an array x and an array y. y is presented below and is values of lumiosity as a function of days (x) for a supernova. I would like to find the proportionality of luminosity to time, L \prop x^(?). Here are my failed attempts using nlinfit. Also, y values are part of another array Lbolos.
if true
x = 960:1236
y = Lbolos(261:537)
model = @(beta, s)( ( Lbolos(261)/960^(beta) )*s^(beta) ) % Lbolos(261)/960^(beta) is for normalisation/start at the right place
initial = [-4]
coeff = nlinfit(x,y, model, initial)
end
if true
x = 960:1236
y = Lbolos(261:537)
model = @(beta, s)( beta(1)*s^(beta(2)) )
initial = [Lbolos(261)/(960^(-4)) -4]
coeff = nlinfit(x,y, model, initial)
end
both times the error message is: Error using nlinfit Error evaluating model function
Help is much appreciated, also a full example if possible<3
if true
y = 1×277
7.436e+41 7.3707e+41 7.3046e+41 7.2376e+41 7.17e+41 7.1017e+41 7.0328e+41 6.9633e+41 6.8933e+41 6.8229e+41 6.752e+41 6.6808e+41 6.6093e+41 6.5374e+41 6.4654e+41 6.3931e+41 6.3207e+41 6.2482e+41 6.1757e+41 6.1031e+41 6.0305e+41 5.958e+41 5.8855e+41 5.8132e+41 5.7411e+41 5.6691e+41 5.5973e+41 5.5258e+41 5.4546e+41 5.3836e+41 5.313e+41 5.2427e+41 5.1729e+41 5.1034e+41 5.0344e+41 4.9658e+41 4.8977e+41 4.8301e+41 4.7631e+41 4.6965e+41 4.6305e+41 4.5651e+41 4.5002e+41 4.436e+41 4.3723e+41 4.3093e+41 4.2469e+41 4.1852e+41 4.1241e+41 4.0636e+41 ...
end

Star Strider

### Star Strider (view profile)

on 18 May 2018
There is usually more to the error than simply: ‘Error using nlinfit Error evaluating model function’
Is it returning ±Inf or NaN values?
Oliver Andersen

### Oliver Andersen (view profile)

on 19 May 2018
It was not returning either of these. The simple fix Torsten gave on adding a dot after the s fixed it! Many thanks for asking though!

### Torsten (view profile)

on 18 May 2018
Edited by Torsten

### Torsten (view profile)

on 18 May 2018

x = 960:1236
y = Lbolos(261:537)
model = @(beta, s)( beta(1)*s.^(beta(2)) )
initial = [Lbolos(261)/(960^(-4)) ; -4]
coeff = nlinfit(x,y, model, initial)
end

Oliver Andersen

### Oliver Andersen (view profile)

on 19 May 2018
Thank you so much! That dot made all the difference.