Improving fitting of curve to data

Hi, I need to fit curve to data, the curve has to be of power-law form with non-constant exponent, which is some function of x (x is the independent variable), the curve form is a*x^(n(x)) where a is constant and n can be any function of x. I use 'fittype' and then 'fit' and get general fit, but it's not good enough, I think that better fit can be achieved with this Eq. form and data. How can I improve the fitting?

Best Regards,

[Matt J]

and n can be any function of x.

You need to decide on a parametric form for n(x). If n(x) can be any function, then effectively no information about the curve is available. Just by choosing n(x)=log_x(f(x))/a, I can make the curve into any f(x) that I wish.

[Roi]

Thanks, I tried different forms including e.g. n=b*x^c or n=b*x^c+d*x (where b, c and d are constants). My question is whether there are some numerical ways to improve the fitting?