dtsien高手請進

非常謝謝[您以下是我的問題log-likelihood:l(theta1

theta2

r)=(y_1 - theta1*y_0)^2*F(y_0 - r) (y_1 - theta2*y_0)^2*(1-F(y_0 - r)) (y_2 - theta1*y_1)^2*F(y_1 - r) (y_2 - theta2*y_1)^2*(1-F(y_1 - r)) . . . (y_n - theta1*y_n-1)^2*F(y_n-1 - r) (y_n - theta2*y_n-1)^2*(1-F(y_n-1 - r))其中F為標準常態分配的CDF經過對theta1 theta2 和 r 偏微之後得到的 normal equation:(y_1 - theta1*y_0)*y_0*F(y_0 - r) (y_2 - theta1*y_1)*y_1*F(y_1 - r) . . . (y_n - theta1*y_n-1)*y_n-1*F(y_n-1 - r)=0(y_1 - theta2*y_0)*y_0*(1-F(y_0 - r)) (y_2 - theta2*y_1)*y_1*(1-F(y_1 - r)) . . . (y_n - theta2*y_n-1)*y_n-1*(1-F(y_n-1 - r))=0[(y_1 - theta1*y_0)^2 - (y_1 - theta2*y_0)^2]*f(y_0 - r) [(y_2 - theta1*y_1)^2 - (y_2 - theta2*y_1)^2]*f(y_1 - r) . . . [(y_n - theta1*y_n-1)^2 - (y_n - theta2*y_n-1)^2]*f(y_n-1 - r)=0其中f為標準常態分配的pdf所以我寫了以下的程式:x=[0.001;-0.032;1.122];syms theta1 theta2 ry1=0;y2=0;y3=0;for i=1:2y1=y1 (x(i 1

1)-theta1*x(i

1))*x(i

1)*(2-erfc((x(i

1)-r)/(2^0.5)))/2;y2=y2 (x(i 1

1)-theta2*x(i

1))*x(i

1)*(1-(2-erfc((x(i

1)-r)/(2^0.5)))/2);y3=y3 ((x(i 1

1)-theta1*x(i

1))^2-(x(i 1

1)-theta2*x(i

1))^2)*exp(-(x(i

1)-r)^2/2);ends=solve(y1

y3);thetahat1=subs(s.theta1)thetahat2=subs(s.theta2)rhat=subs(s.r)其中(2-erfc((x(i

1)-r)/(2^0.5)))/2 就是我寫出來的CDF而erfc是內建程式但是跑不出來不好意思寫了這麼長麻煩您了
Dear Sam:MLE is for model parameters estimation based on sampling data. Now you have 3 data points (x=[0.001;-0.032;1.122]) and try to