Estimation of transition rates in dynamic household models.
Estimation of transition rates in dynamic household models.
复制标题
动态家庭模型中的转变率估计。
DOI：
--
复制
发表时间：
1988
期刊：
影响因子：
--
通讯作者：
É. Lelièvre
中科院分区：
文献类型：
--
作者：
D. Courgeau;É. Lelièvre
本章讨论非参数半参数和参数模型中瞬时转移率的估计。的双变量和多变量的问题进行了介绍和讨论，使用的数据从回顾性的生活史调查在法国于1981年。这一探索指出了未来研究的一些重要问题。一个重要的问题是，不可能估计完整的模型，因为这将需要一个巨大的样本，每个单元中有许多观察结果。没有这方面的详细资料。因此，作者主要根据人口数据的早期经验对主要影响进行了一些假设。例如，这适用于自上次事件以来的持续时间的显著效应，而不是自先前事件以来的持续时间。另一个问题是在这些分析中引入更多的时间依赖性解释变量。当重新审视双变量问题时，作者引入了这样一个时间依赖变量，即在时间t是否结婚。这种做法需要推广。第三个问题在前面的分析中没有清楚地出现，但似乎对未来的研究非常重要。作者引入了以前不同的解释变量来控制它们的效果。然而，被研究个体的其他特征干扰了它们，而不可能测量这些特征。作者可以将这种未观察到的异质性纳入模型，但对这个问题的研究才刚刚开始。最后，作者介绍了一个“模糊时间”，导致非常重要的问题是难以处理。转换率的估计需要一个非常精确的时间尺度上的事件可以放置。然而，在研究人口统计事件时，作者没有观察到这样的时间尺度，而是观察到一个更“模糊”的时间。在6个月甚至一年的时间间隔内工作时，知道出生或迁移是否首先发生似乎并不重要。
This chapter deals with the estimation of instantaneous transition rates in nonparametric semiparametric and parametric models. The bivariate and multivariate problems are introduced and discussed using data from a retrospective life history survey undertaken in France in 1981. This exploration pinpoints some important issues for future research. An important problem is that it was not possible to estimate complete models because this would require a huge sample with many observations in each cell. Such detailed information was not available. Therefore the authors introduced some assumptions on the main effects mainly based on earlier experiences with demographic data. For example this applies to the prominent effect of the duration since the last event rather than to the duration since earlier ones. Another problem will be to introduce more time-dependent explanatory variables into these analyses. When revisiting the bivariate problem the authors introduce such a time-dependent variable namely being married or not at time t. Such an approach needs to be generalized. A 3rd problem did not appear clearly from the previous analyses but seems very important for future research. The authors introduced previously different explanatory variables to control their effect. However other characteristics of the studied individuals interfere with them while there is no possibility of measuring these characteristics. The authors can incorporate this unobserved heterogeneity into the models but research on this problem has only just begun. Lastly the authors have introduced a "fuzzy time" that leads to very important problems which are difficult to deal with. The estimation of transition rates needs a very precise time scale on which events can be placed. However when studying demographic events the authors did not observe such a time scale but rather a more "fuzzy" time. It does not seem important when working with a time interval of 6 months or even a year to know whether a birth or a migration occurred 1st.