I have a problem with minimizing a likelihood function(LLF). I wrote everything and my code works without any error. but my problem is the scipy. minimize only change 1 parameter and the other ones remain the same as the initial inputs. (i changed the method but it didn't work)
this is my LLf:
and these are the LLf parameters:
i wrote each parameter in a seperate function and LLF is the objective function. here you can see my code:
def Rt(r1):
return (np.ones([n, n])) * r1 + (np.eye(n)) * (1 - r1)
def absoluteRt(r1):
return ((1 - r1) ** (n - 1)) * (1 + (n - 1) * r1)
def inverseRt(r1):
return (np.ones([n, n]) * (1 / (1 - r1))) - ((np.eye(n)) * (r1 / ((1 - r1) * (1 + (n - 1) * r1))))
def Dt(t1):
return np.diag(np.asarray(ht.iloc[:, t1]))
def ytilde(t1):
return np.dot((Dt(t1)), (np.subtract(np.asarray(returns.iloc[:-1, t1]), (np.subtract(np.asarray(m), (
np.multiply(np.asarray(phi.iloc[:, 0]), np.asarray(returns.iloc[:-1, t1 - 1]))))))).T)
def constraint1(r1):
return np.abs(r1) - epsilon
def objective(r1):
for i in range(t):
z = np.log(absoluteRt(r1[i])) + np.dot(ytilde(i).T, (np.dot(inverseRt(r1[i]), ytilde(i))))
return (1) / 2 * z
w = [0.1 , 0.5, 0.5, 0.5]
con1 = {'type': 'ineq', 'fun': constraint1}
cons = ([con1])
b = (((-1)/131)+epsilon , 1-epsilon)
bnds=tuple( b for e in range(t))
solution= minimize(objective,w,method='SLSQP', constraints=cons , bounds=bnds)
and the result is:
fun: 23.564446928934615
jac: array([ 0. , 0. , 0. , 17.42714119])
message: 'Optimization terminated successfully.'
nfev: 12
nit: 2
njev: 2
status: 0
success: True
x: array([ 1.00000000e-01, 5.00000000e-01, 5.00000000e-01, -2.44142484e-11])
where is my mistake? what am I doing wrong?
