Prevention researchers often analyze nonnegative continuous outcomes that are skewed and contain many zero observations. Variables with these features can arise when investigators measure low-base-rate behaviors (e.g., bullying, substance use) or use measurement tools that cannot differentiate individuals with low standing on the target construct. Common analytic procedures such as linear regression are often inappropriate for such data and may result in biased results. This article provides a tutorial on alternative methods that may be more suitable for the analysis of nonnormal, zero-bounded continuous outcomes. The meaning of the zero values is an important consideration when choosing an analytic approach. We discuss 2 common causes of abundant zeros (i.e., censoring due to measurement limits and true absence of the outcome), highlight problems that arise with the application of standard statistical methods to such outcomes, and describe promising but less well-known analytic alternatives. We illustrate these techniques using data from a study of bullying prevention in urban youth. Two approaches, Tobit regression and two-part models, are particularly useful methods for handling skewed nonnegative outcomes with several zero values. Familiarity with the issues and techniques we present may help researchers to make more informed analytic choices when confronted with such outcomes.

中文翻译：

---

分析带有多个零位的偏斜的连续结果：社会工作和青年预防科学研究者的教程

预防研究人员经常分析偏斜的非阴性连续结果，其中包含许多零观测值。当研究人员测量低基率行为（例如，欺凌，滥用毒品）或使用无法区分目标构筑率低的个人的测量工具时，可能会出现具有这些特征的变量。诸如线性回归之类的常用分析程序通常不适用于此类数据，并且可能导致结果有偏差。本文提供了有关替代方法的指南，这些方法可能更适合于分析非正常，零界连续结果。零值的含义是选择分析方法时的重要考虑因素。我们讨论了两个导致大量零的常见原因（例如，由于测量限制和结果的真实缺失而进行的审查），突出显示了将标准统计方法应用于此类结果时出现的问题，并描述了有前途但鲜为人知的分析方法。我们使用来自对城市青年中的欺凌预防研究的数据来说明这些技术。两种方法，Tobit回归和两部分模型，对于处理带有多个零值的偏斜非负结果特别有用。面对此类结果，熟悉我们提出的问题和技术可以帮助研究人员做出更明智的分析选择。是处理具有多个零值的偏斜非负结果的特别有用的方法。面对此类结果，熟悉我们提出的问题和技术可以帮助研究人员做出更明智的分析选择。是处理具有多个零值的偏斜非负结果的特别有用的方法。面对此类结果，熟悉我们提出的问题和技术可以帮助研究人员做出更明智的分析选择。