Randomistas: How Radical Researchers Are Changing Our World.
Randomistas: How Radical Researchers Are Changing Our World.
复制标题
随机主义者：激进的研究人员如何改变我们的世界。
DOI：
10.1080/00031305.2019.1676111
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Gary Saretsky
中科院分区：
文献类型：
--
作者：
Gary Saretsky
代码在做什么以及脚本中的描述量因章节而异。重要的是要澄清，这本书确实解释了每一章中使用的代码，这是非常有益的。然而，在脚本文件中有更多的描述将对读者更有用，并将减少读者目前有义务在软件和书籍之间来回导航的数量。解决这个问题的一种可能方法是使用R Markdown文件。在R Markdown文件中，注释很容易以可视化的方式与代码区分开来，使文档更有组织，更容易阅读。在R Markdown文件中添加每个分析的简要介绍和代码的额外描述也有助于更好地理解代码本身和整个分析，并有助于重现性。此外，使用R Markdown文件，读者可以很容易地在同一位置看到代码和输出之间的关系，为学习R语言的人提供了一个清晰的好处。在软件包方面，作者使用现有的R软件包，如psych和lavaan进行因子分析，mirt用于一维和多维IRT模型，equate用于非基于IRT的等值。该软件包还用于创建交互式图，其中一个可以用作教学工具，以帮助读者看到改变项目难度，项目区分度和猜测对项目反应函数的影响。在每一章的结尾，作者经常提到其他没有使用但对读者有帮助的R包。除了使用现有的R软件包外，作者还开发了一个名为hemp的新软件包（目前可在GitHub上获得），其中包含本书中使用的所有数据集以及主要用于CTT和概化理论章节的一般心理测量学函数。本书的项目分析部分使用基本的R函数解决项目难度和项目区分度（通过点二阶相关性测量），并从大麻包中引入独特的函数来计算项目区分度指数，项目可靠性指数，项目有效性指数和干扰物分析。在阅读本章之后，读者将拥有进行项目分析所需的R函数;然而，使用这些函数假设所使用的数据已经被评分为0和1。由于本书假设没有R的先验知识，因此新手R用户将受益于如何重新编码他们的数据或引用的解释。如果hemp包有一个创建项目分析报告的整体功能，那么它也会非常有用和实用，然后可以导出为PDF/HTML文件。除了上面提到的基本CTT函数外，hemp软件包还在概化理论的背景下引入了新的函数，以估计概化和决策研究中的方差分量，概化和可靠性系数。在这本书出版的同一年，Mair的现代心理测量学与R（2018）也出版了。两本书所涵盖的内容之间存在重叠（例如，CTT，一些因素分析技术和一些IRT模型）;然而，Mair的书涵盖了更多的方法，如结构方程模型，偏好建模，对应分析和时间序列等方法。这两本书之间的一个重要区别是，梅尔假设读者已经熟悉R，而如上所述，Desjardins和Bulut的书没有这样的假设。总的来说，Desjardins和Bulut的《使用R的教育测量和心理测量手册》是教育测量和心理测量领域的一个独特和急需的补充。对于不熟悉R并且需要对教育测量和心理测量分析技术进行用户友好介绍的教授、研究人员和研究生来说，这本书将是一个很好的资源。
what the code is doing and the amount of description in the scripts varies from chapter to chapter. It is important to clarify that the book does explain the code used in each chapter, which is very beneficial. However, having more descriptions in the script files would be even more useful to readers and would decrease the amount of back and forth a reader is currently obligated to navigate between the software and the book. One possible approach to fixing this would be to use R Markdown file(s). In an R Markdown file, comments are easily distinguished from the codes in a visual way, making the document much more organized and easier to read. Adding brief introductions for each analysis and additional descriptions of the codes in the R Markdown file would also contribute to a greater understanding of the codes themselves, and of the analysis as a whole, and would facilitate reproducibility. In addition, with an R Markdown file, the reader could easily see the relationship between the code and output in the same location, providing a clarifying benefit to those learning the R language. In terms of the packages, the authors use existing R packages such as psych and lavaan for factor analysis, mirt for unidimensional and multidimensional IRT models, and equate for non-IRT-based equating. Theshiny package is also used to create interactive plots, one of which can be used as a teaching tool to help readers see the effect of changing item difficulty, item discrimination, and guessing on the item response function. At the end of each chapter, the authors often refer to other R packages that were not used but could be helpful to readers. In addition to using the existing R packages, the authors developed a new package called hemp (currently available on GitHub) which contains all the datasets used in the book and general psychometrics functions primarily used in the chapters covering CTT and generalizability theory. The item analysis section of the book addresses item difficulty and item discrimination (as measured by the point biserial correlation) using basic R functions and introduces unique functions from the hemp package to calculate item discrimination index, item-reliability index, item-validity index, and distractor analysis. After reading this chapter, readers would have the R functions needed to conduct an item analysis; however, the use of these functions assumes that the data used is already scored as 0’s and 1’s. Since this book assumes no prior knowledge of R, novice R users would benefit from an explanation of how to recode their data or references. It would also have been very useful and practical if the hemp package had an overall function that creates an item analysis report, which could then be exported as a PDF/HTML file. In addition to the basic CTT functions mentioned above, the hemp package also introduces new functions in the context of generalizability theory to estimate variance components, generalizability and dependability coefficients in generalizability and decision studies. In the same year of the publication of this book, Modern Psychometrics with R by Mair (2018) was also published. There is an overlap between the content covered by both books (e.g., CTT, some factor analytic technics, and some IRT models); however, Mair’s book covers more methods such as structural equation models, preference modeling, correspondence analysis, and time series, among other methods. An important difference between the two books is that Mair assumes the reader already has familiarity with R while, as mentioned above, Desjardins and Bulut’s book makes no such assumption. Overall, the Handbook of Educational Measurement and Psychometrics Using R by Desjardins and Bulut is a unique and much-needed addition to the field of educational measurement and psychometrics. This book would be a great resource for professors, researchers, and graduate students who are not familiar with R and need a user-friendly introduction to educational measurement and psychometric analysis techniques.