Departmental Bulletin Paper 潜在変数を含む統計モデルにおける効率的なパラメータ推定

中村, 永友  ,  土屋, 高宏

We propose an efficient bootstrap method for statistical models which have the latent variables as a weight for each data point. The proposed method, can also be seen as the resampling of a way to fully reflect the structure of the estimated statistical model f(θ^^Λ)from the observed data. Examples of such statistical models are the t-distribution model, the M-estimator, and the finite normal mixture model, etc. The effectiveness of the proposed method is verified through numerical experiments.

Number of accesses :  

Other information