论文部分内容阅读
在社会心理学和行为学的研究中,记录某些健康或行为结果发生频率的计数中(如在一段时间内无防护措施的性行为的次数)往往含有大量的零,这是因为当某些对象对于某种研究行为没有危险时就会产生“结构性零”。不像随机零(结果可以是大于零,但是也可能由于样本变异性而成为零),结构性零在统计和临床上通常是非常不同的。如果两种类型零的差异被忽略,就可能会导致对结果和研究发现的错误解释。然而在实践中,结构性零经常会没有被观察到而这种潜在性使数据分析复杂化了。在这篇文章中,我们专注于一种模式,即通常用于解决零膨胀数据的零膨胀泊松(Zero-inflated Poisson,ZIP)回归模型。首先,我们对结构性零和ZIP模型做一个简要概述。然后我们以一项青春期少女艾滋病高危性行为的研究数据来阐述ZIP模型。文中还附有SAS和Stata的示例代码,以帮助运行和解释ZIP分析。
In studies of social psychology and behavior, the frequency of recording some health or behavioral outcomes (such as the number of unprotected sexual acts over time) tends to contain a large number of zeros because when some An object produces “structural zero ” when it is not at risk for a research activity. Unlike random zero (the result can be greater than zero, but it can also be zero due to sample variability), structural zero is often very different statistically and clinically. If the difference between the two types of zero is ignored, it may lead to misinterpretation of the findings and research findings. In practice, however, structural zero is often not observed and this potential complicates data analysis. In this article, we focus on one model, a Zero-inflated Poisson (ZIP) regression model commonly used to solve zero-expansion data. First, let’s take a brief overview of the structural zero-sum ZIP model. Then we describe the ZIP model with data from a study of adolescent girls at high risk of HIV / AIDS. The article also comes with sample code for SAS and Stata to help run and interpret ZIP analysis.