統計学輪講（第３２回）

日時　　  2001年11月27日（火）　15時00分〜15時50分
場所　　  経済学部5階視聴覚室
講演者　  李　建平(経済M2)
演題　    Weibull Proportional Hazard Model and Comparison 
          between MLE and LSE

概要：  
  Central concept in a variety of situations lies not in
unconditional probabilities but in conditional ones. Hazard model, (or
hazard function, hazard rate) is a convenient approach for duration data
analysis. In this report, K-M nonparametric method, parametric method and
Cox's proportional hazard model in terms of hazard functions will be
introduced first. Major attention is restricted to Weibull proportional
hazard model. Compared with the usual maximum likelihood estimation (MLE) of
a proportional hazard model, the least squares estimation (LSE) under the
log-linear formulation is investigated thoroughly. Several examples are also
provided to illustrate how the LSE behaves differently from MLE for data
without censoring and censored data respectively.