I write the following code for AR(2) process maximum likelihood estimation
global y t;
y = [1 2 4 6 4 12 11 13 11 14 16 17]';
T = length(y);
t = [1:T]';
x0 = [1, 1, 0.1, 0.1];
[x, loglval] = fminunc(@(x) logl(x(1), x(2), x(3), x(4)), x0);
disp(x)
and the log likelihood function
function val = logl(a, b, p, s)
global y t T
f = y(2:T) - a*ones(T-1,1) - b*t(2:T) - p*y(1:T-1) - p*a*ones(T-1,1) - p*b*t(1:T-1);
val = -log(2*pi)*T/2 - T*log(s)/2 + log((1-p^2)^(1/2)) - (1/(2*s))*((1-p^2)*(y(1)-a-b*t(1))^2 + sum(f.*f));
end
But I get Supplied objective function must return a scalar value error. ı do not understand the error. I run the code with isscalar(val) and it returns 1. Could you help me?
