On the Generalized Maximum Likelihood Prediction
Teruaki Ishikawa, Takanori Tamiya, Kouichi Kotani · Institutional Repositories DataBase (IRDB) · 1994
In this paper, we discuss the properties of maximum likelihood predictors (MLPs) for a future random variable based on past samples. For that purpose, we give two theorems and some examples. Furthermore we propose a principle of generalized maximum likelihood giving a new prediction method. By illustrating examples, it is shown that this new method gives some usefull predictors for some problems.