對於擁有潛在類別的迴歸模型，我們通常都用的EM 演
算法來做參數估計。而近來模糊類別最大概似法(Fuzzy
Classification Maximum Likekihood)；簡稱FCML)在具潛
在類別卜瓦松迴歸的參數估計上有不錯的成效。因此本文
中，我們將進一步探討在具潛在類別的負二項迴歸模型的參
數估計問題，這兩種演算法的精確度與效率。透過數值模擬
的結果顯示，FCML演算法在此模型中，僅在特定的模型假設
下有較好的結果。整體而言，與EM演算法相較，互有優劣。

Regarding to Regression Model, which possesses Latent
class model, we usually use EM algorithm to calculate Parameter
Eastimations. Recently, Fuzzy Classification Maximum
Likelihood (abbreviated as FCML) has good outcome in Poisson
Regression with possession of Latent class model. According to
this result, we go a step further on discussion of the
efficiency and the accuracy of negative binomial regression
analysis with possession of Latent class model. By the result
of numerical imitation, FCML algorithm has only good outcome
in specifically hypothesized model. From the whole of
comparison, both FCML algorithm and EM algorithm have
advantages and disadvantages.

[1] D.B Rubin(1977)，A.P Dempster and N.M Laird ，Maximum Likelihood form
incomplete data via the EM algorithm(with discussion)，Journal of the Royal Statistical
B39:1-38﹒
[2] M.S.Yang(1993)，On a class of fuzzy classification Maximum likelihood
procedures，Fuzzy Sets and Systems 57:365-375﹒
[3] M.S.Yang、C.Y.Lai(2005) ，Mixture Poisson regression models for heterogeneous
count data based on latent and fuzzy class analysis﹒
[4] L.A.Zadeh(1965)，Fuzzy sets，Information and control 8:338-353﹒
[5] J.T Grogger(1990)，The deterrent effect of capital punishment:an analysis of daily
homicide counts，Journal of the American Statistical Association 85:295-303﹒
[6]M.Wedel，W.S.DeSarbo，J.R. Bult and V.Ramaswamy，A latent class Poisson
Regression model for heterogeneous count data，Journal of Applied Econometrics 8
(1993)397- 411