|
PureBytes Links
Trading Reference Links
|
I'm not sure that I have your answer, because I don't have the background to make you a mathematical demonstration, but I quickly read the document and
I think that your answer is between the pages 11 and 12 ...
"When the projection errors in the measurment equation are not Gaussian, MLE is more involved ... "
For me "involved" in this case means "complicated".
And the next part clarify your problem.
Raf
15.11.01 19:06:26, "Bilo Selhi" <biloselhi@xxxxxxxxxxx> a écrit:
>yes, i understand that.
>here is the paper that i used as reference and it mentions
>http://wrds1.wharton.upenn.edu/fic/wfic/papers/00/p0028.html
>on page 6,7,8 that they use kalman filter as quasi-maximum likelihood
>estimate for those 3 parameters.
>i think the reason why they do that is because kalman is simpler?
>why not just use straight MLE estimator in this case?
>i am having trouble understanding the use of kalman in this case,
>can you clarify?
>bilo.
>
>
>----- Original Message -----
>From: "^^_^^" <ferrue@xxxxxxxxxx>
>To: <omega-list@xxxxxxxxxx>
>Sent: Thursday, November 15, 2001 6:58 AM
>Subject: Re: adaptive entries, exits. MLE, QMLE est. in Stoch. Vol. models
>
>
>> Bilo,
>>
>> If you have only one unknow then simply rewrite the equation, else you can
>use an optimization algo with some constraints.
>>
>> Otherwise I didn't understand your question ...
>>
>> HTH
>> Raf
>>
>>
>>
>>
>
|