Quadratic Artificial Likelihood Functions Using Estimating Functions

摘要: Abstract.  A vector-valued estimating function, such as the quasi-score, is typically not gradient of any objective function. Consequently, an analogue likelihood function cannot be unambiguously defined by integrating This paper studies inference in framework optimal functions. We propose a quadratic artificial for The uniquely identified potential from vector field decomposition imposing some natural restriction on divergence-free part. shown to resemble genuine number respects. bootstrap version also studied, which may used selecting root estimate among multiple roots equation.