Valencia-Orozco Andrea, Parra-Lara Luis G, Martínez José W, Tovar-Cuevas José R
Escuela de Estadística, Facultad de Ingeniería, Universidad del Valle. Cali, Colombia.
Centro de Investigaciones Clínicas, Fundación Valle del Lili. Cali, Colombia.
Rev Peru Med Exp Salud Publica. 2019 Apr-Jun;36(2):341-348. doi: 10.17843/rpmesp.2019.362.4269. Epub 2019 Aug 22.
This article describes a methodology that allows an approach to alternative right-censored probabilistic models for the analysis of survival, different to those usually studied (exponential, gamma, Weibull, and log-normal distribution) since it is possible that the data do not always fit with sufficient precision due to existing distributions. The methodology used allows for greater flexibility when modeling extreme observations, generally located in the right tail of data distribution, which admits that some events still have the probability of occurring, which is not the case with traditional models and the Kaplan-Meier estimator, which estimates for the longest times, survival probabilities approximately equal to zero. To show the usefulness of the methodological proposal, we considered an application with real data that relates survival times of patients with colon cancer (CC).
本文描述了一种方法，该方法允许采用替代的右删失概率模型来分析生存情况，这与通常研究的模型（指数分布、伽马分布、威布尔分布和对数正态分布）不同，因为由于现有分布的原因，数据可能并不总是能以足够的精度拟合。所使用的方法在对极端观测值进行建模时具有更大的灵活性，这些极端观测值通常位于数据分布的右尾，这意味着某些事件仍有发生的概率，而传统模型和卡普兰 - 迈耶估计器并非如此，后者对最长时间的生存概率估计近似为零。为了展示该方法建议的实用性，我们考虑了一个使用结肠癌（CC）患者生存时间真实数据的应用。
