ILIAD data - cleaned

A secondary data analysis was conducted to using data from the Independent Lexical Instruction and Development (ILIAD) project, a supplementary vocabulary program that taught 377 sophisticated vocabulary words across three years to 350 first, second, and third grade students (Goldstein et al., 2017). Two outcome measures were used to assess word learning, a decontextualized definition task and an expressive labeling task. A word was considered learned when children defined and/or labeled vocabulary targets, respectively. The target vocabulary words were characterized for analysis based on available database estimates of their individual word frequency, age of acquisition, phonotactic probability, neighborhood density, and level of concreteness. First, we compare how well the models fit the data using the full ILIAD dataset (first, second, and third, grades combined). While we want to highlight the differences in modeling techniques for each individual grade level, it would be exhaustive. We randomly selected data from first grade to illustrate the unique differences between MARS, multiple regression, and stepwise regression. This data contains lexical characteristics and decontextualized and expressive learning outcomes for 143 vocabulary words.

本研究开展了一项二手数据分析，所用数据来自独立词汇教学与发展（Independent Lexical Instruction and Development, ILIAD）项目，该项目为一项补充性词汇教学计划，在三年周期内为350名一、二、三年级学生讲授了377个高阶词汇（Goldstein等人，2017）。本研究采用两项结果指标评估词汇学习成效：其一为脱离语境定义任务，其二为表达性命名任务。当儿童能够正确定义目标词汇，或/且能够正确对目标词汇进行命名时，即判定该儿童已掌握对应词汇。研究将基于现有数据库给出的各词汇个体词频、习得年龄、音位配列概率、词汇邻接密度以及具体性程度等估计值，对目标词汇进行特征标注以用于后续分析。首先，本研究将使用完整的ILIAD数据集（合并一、二、三年级数据），对比不同模型对数据的拟合效果。尽管我们希望凸显不同建模技术针对各年级单独分析的差异，但此类操作将过于繁琐。因此，我们随机选取一年级学生的数据集，用以展示MARS、多元回归以及逐步回归三种方法间的独特差异。该数据集包含143个目标词汇的词汇特征数据，以及脱离语境定义任务与表达性命名任务的学习结果数据。