CAJ | 학술논문

使用Poisson分布刻画数据，要求样本数据不为0且其方差和均值相等。然而在很多实际问题中，计数数据很难满足上述要求。针对这一现象，文章研究了零截尾广义Poisson回归模型，基于局部影响分析方法，在3种不同扰动模式下得到影响曲率表达式及强影响点，并通过实例说明了该模型和方法的有效性。
사용Poisson분포각화수거，요구양본수거불위0차기방차화균치상등。연이재흔다실제문제중，계수수거흔난만족상술요구。침대저일현상，문장연구료령절미엄의Poisson회귀모형，기우국부영향분석방법，재3충불동우동모식하득도영향곡솔표체식급강영향점，병통과실례설명료해모형화방법적유효성。
In real application, it is required the equality of the sample mean and variance, which is difficult to be satisified for many real data analysis. Meanwhile, it is often found in some situations that count data not only has no zero but also structurally excludes having 0 counts. In this paper, the zero-truncated generalized Poisson regression was investigated, and the local influence analysis under three different disturbance modes based on curvature mea-surement method was mainly discussed. Finally, the real data analysis was used to illustrate the usefullness pf the proposed model and approach.