Interactive maximum likelihood estimation
Autor: | D. L. Prager, Peter E. Wellstead |
---|---|
Rok vydání: | 1980 |
Předmět: | |
Zdroj: | International Journal of Control. 32:1005-1030 |
ISSN: | 1366-5820 0020-7179 |
Popis: | This paper describes an approach to maximum likelihood estimation which is based upon recursive hypothesis testing. The advantage of the method over existing hill-climbing and approximate recursive methods lies in the degree of information which can be obtained from the algorithm. In particular, cross-sections through the loss function are provided in a graphical form. This allows the analyst to see the likelihood function evolve, and make subjective decisions based upon this information. Effectively, the method computes all the relevant points on the loss function surface, making the technique ideal for short data runs where conditioning of the likelihood function is a problem. The method compares favourably with alternative maximum Likelihood algorithms in terms of computational effort and has none of their convergence problems. |
Databáze: | OpenAIRE |
Externí odkaz: |