Statistical modelling of skewed and semi-continuous data in randomised controlled trials (RCTs): An application of two-part models in RCTs of chronic

慢性疼痛是指持续超过正常愈合时间的疼痛，即使找不到明显的原因也会发生，据估计，英国有1400万人患有慢性疼痛，随着老年人口的增加，全球与疼痛相关的疾病的负担越来越大。使用不适当的分析技术严重影响随机对照试验的结果，往好了说，将降低效果估计的精确度；往坏了说，结果可能极具误导性。两部分模型是一种灵活的建模框架，经常用于计量经济学中具有大量零值和非零值的高度倾斜或非正态的数据。这项研究的目标是开发和评估两部分模型在测量疼痛和疼痛相关残疾的结果的背景下在肌肉骨骼RCT中的应用。两部分模型的性能将通过将这些模型与模拟数据相匹配来探索，以充分评估治疗效果估计的偏差和准确性，并将通过生成模拟数据集来评估它们的性能。两部分模型的灵活性增加有望导致更好地拟合疼痛数据的模型，因此，最重要的是，更可靠的推断。

Chronic pain is pain that persists beyond the normal time of healing, and can occur even when no obvious cause can be found with estimates suggesting 14 million people live with chronic pain in the UK and the global burden of pain related conditions increasing with an increasing elderly population. The use of inappropriate analytical techniques seriously affects the findings of RCTs and will at best the precision of effect estimates will be reduced; at worst results might be grossly misleading. Two-part models are a flexible modelling framework often used in econometrics for data with a large number of zeroes and non-zero values that are highly skewed or non-normal. The goal of the research will be to develop and assess the application of two-part models in the context of outcomes that measure pain, and pain related disability, in musculoskeletal RCTs.Performance of two-part models will be explored by fitting these models to simulated data to fully assess the bias and accuracy of estimates of treatment effects and their performance will be assessed by generating simulated datasets. The increased flexibility of two-part models is expected to lead to better fitting models for pain data and thereby, most importantly, more reliable inference.