Truncated survival data and Bayesian optimal design for changepoint problems

我的第一个研究领域是生存分析。 假设观察是随着时间的推移进行的，对象（例如，可以是患者）具有明确定义的开始时间和明确定义的终点。例如，起始时间可以是痴呆的发病日期，终点可以是死亡。 兴趣在于对所谓的“生存时间”进行统计推断-从开始到终点;例如，患者患有痴呆症的时间长度。生存分析的特点是某些项目的生存时间观察不完全。发生这种情况的最常见原因是，在研究结束时关注的事件尚未发生，或者受试者在事件发生前失访。如果发生这种情况，生存时间被称为右删失。然而，有时由于研究设计的原因，生存时间短的受试者根本没有被观察到作为研究的一部分。然后，生存时间被称为左截断。我擅长分析右删失左截断的生存数据。 我的第二个研究领域是变点问题的优化设计。按项目顺序进行观察（例如，治疗前后患者的血压读数）。 怀疑该治疗可能对一定比例的受试者有效。然而，如果治疗有效，不知道治疗后多久会出现这种效果。也就是说，会有预期的变化，但变化之前的时间间隔是未知的。我研究观察的最佳时间，以最好地估计治疗效果，并估计治疗有效的受试者比例。

My first area of research is in the area of survival analysis. Suppose that observations are taken over time, on items (they could be patients, for example) with a well-defined initiation time and a well defined end point. For example, the initiation time could be the onset date of dementia and the endpoint could be death. Interest lies in making statistical inference about the so-called "survival time"-between initiation and the endpoint; for example, the length of time that patients live with dementia. The feature of survival analysis is that some of the survival times for the items are incompletely observed. Most often this happens because either the event of interest has not yet occurred by the time the study ends or the subjects has been lost to follow-up before the event has occurred. If this happens the survival times are said to be right censored. Sometimes however, as a result of the study design, subjects who have short survival times are simply not observed as part of the study. The survival times are then said to be left truncated. I specialize in analyzing survival data that are right censored and left truncated. My second area of research is optimal design for changepoint problems. Observations are to be taken sequentially on items (e.g. blood pressure readings on patients before-and-after a treatment). It is suspected that the treatment might be effective on a certain proportion of subjects. However, if the treatment is effective it is not known how long after the treatment this effect will occur. That is, there will be an anticipated change but the time lag before the change is unknown. I work on the optimal timing of the observations to best estimate the treatment effect and to estimate the proportion of subjects for which the treatment will be effective.