Workshop on Categorical Data Analysis

分类数据被收集在许多应用程序中。随着计算资源和计算能力的广泛使用，一套全面的分类数据方法已经开始出现。然而，重要的公开研究问题依然存在。随着收集的分类数据集越来越大，分析高维数据的挑战变得更加常见。对于分类数据来说，评估相关性结构并对其进行适当的建模比对于正态分布的数据更有问题。同样，当缺失数据不可忽略时，广义估计方程通常会导致有偏估计。分类数据的混合模型和平滑技术是研究的重要领域。贝叶斯方法在模型平均和模型选择过程中变得流行起来。现在特别感兴趣的一个领域是贝叶斯诊断学(例如，残差和后验预测概率)的发展，它是拟合模型的副产品。在这次研讨会上，分类方法论领域的研究领袖和年轻研究人员将汇聚一堂。11位领导人将在研究的前沿发表演讲。年轻的研究人员将有机会展示他们的研究海报，并与经验丰富的研究人员进行广泛的访问。大量的应用程序会产生分类数据，例如性别、种族和教育水平。许多传统的统计方法假设数据来自正态分布，不适用于分类数据。随着计算资源的可用性和计算能力的提高，已经发展了一套全面的分类统计方法。然而，还需要其他的统计方法。在遗传学中，从每个个体收集了大量的分类反应，导致了复杂的分析问题。在许多研究中，一些数据缺失。这可能是因为一个人退出了一项研究，拒绝回答等。有时丢失的数据是随机发生的，这种丢失可能会被忽略。在其他时候，缺失的数据不是随机的，比如当一种治疗让人们感觉很糟糕时，他们就退出了。这种缺失是不容忽视的。这些只是研讨会期间要探索的两个研究领域。讲习班将提供一个极好的机会，讨论分类方法的许多最新重大发展，并确定重要问题和新的研究方向。本研讨会将确定在由新技术驱动的应用领域中似乎最有效的方法。

Categorical data are collected in numerous applications. As computing resources and power have become widely available, a comprehensive set of categorical data methods has begun to emerge. Yet, important open research questions remain. As increasingly large categorical data sets are collected, challenges of analyzing high dimensional data become more common. Assessing the correlation structure and properly modeling it are more problematic for categorical data than for normally distributed data. Similarly, when missing data are non-ignorable, generalized estimating equations usually result in biased estimators. Mixed models and smoothing techniques for categorical data are important areas of research. Bayesian methods have become popular for model averaging and model selection procedures. An area of particular interest now is the development of Bayesian diagnostics (e.g., residuals and posterior predictive probabilities) that are a by-product of fitting a model. In this workshop, research leaders and young researchers in the field of categorical methodology will be brought together. Eleven leaders will present lectures at the cutting-edge of research. Young researchers will have the opportunity to present posters show-casing their research and to visit extensively with seasoned researchers.Numerous applications give rise to categorical data, e.g. sex, race, and education level. Many of the traditional statistical methods assume that the data are from a normal distribution and are not appropriate for categorical data. With the increased availability and power of computational resources, a comprehensive set of statistical methods for categorical has been developed. However, additional statistical methods are needed. In genetics, large numbers of categorical responses are collected from each individual, leading to complex analytical questions. In many studies, some data are missing. This could arise because a person drops out of a study, refuses to respond, etc. Sometimes the missing data occur randomly, and the missingness may be ignored. At other times, the missing data is not at random, such as when a treatment makes people feel badly and they drop out. Such missingness cannot be ignored. These are but two research areas to be explored during the workshop. The workshop will provide an excellent opportunity to discuss the many recent significant developments in categorical methodology and to identify important problems and new research directions. This workshop will identify the methods that seem to work best in areas of application driven by new technology.