You have what, 13 data points? A model with many exponents in it? A model that is surely based on no real physical meaning of the terms in the model?
Expect numerical problems. You will need good starting values.
Worse, z == 0. z is IDENTICALLY zero. Yet you then expect to estimate the power z, z^c6? SIGH.
I can confindently predict that c6 = 17. Oh, wait, I remember, all parameters have the value 42, at least when any possible value will suffice.
The point is, you cannot intelligently estimate these parameters, because there is no unique solution. Any value of c6 will work.
That means your model reduces to
Q = ((c1-c2*ye)*(asin(ye)^c3)+c4*ye^(c5+1))^c7
It is still a mess of parameters put there just to get enough flexibility to fit your data.
So now, lets plot your data, something I should have done before anything else.
You are kidding me, right? A simple low order polynomial model is entirely adequate. Is there sufficient information content in those 13 data points to estimate the parameters tou wish to fit? From somewhere in the distance, you hear a deep, heartfelt sigh, even a groan. Is there any physical reason why you think this model is appropriate, that those parameters have any physical meaning in context?
If you prefer, a simple power curve will do quite well.
>> mdl = fittype('a + b*abs(ye-c)^d','indep','ye')
mdl(a,b,c,d,ye) = a + b*abs(ye-c)^d
>> fittedmodel = fit(ye,Q,mdl,'start',[.01,.1,-.1,1.3])
fittedmodel(ye) = a + b*abs(ye-c)^d
Coefficients (with 95
a = 0.04029 (0.03285, 0.04772)
b = 1.508 (1.45, 1.566)
c = 0.1636 (0.1464, 0.1808)
d = 1.297 (1.227, 1.366)
>> hold on
And that fits quite well, with a far simpler model.