14 Nonparametric maximum likelihood methods

摘要: Publisher Summary This chapter discusses nonparametric maximum likelihood methods. The (NPML) method is a direct attack, via the principle, on problem of dealing with an unknown distribution function in estimation or testing. describes application this approach two distinct but related areas: (1) semiparametric which because incomplete data, underlying has to be estimated together parameters model. improves arbitrarily specifying form that generally not robust against misspecification function. NPML one estimation. . In empirical methods for confidence regions, complete data and conventional M -estimator available parameter objective obtain regions better small sample properties than those obtained by bootstrap asymptotic corrections central limit theorem.