The importance of test equating and its methods in Foreigh language education: test equating as a part of validity study

在当前研究进行的测试环境中，测试项目在每份测试表格的管理后立即向受试者公开。本研究展示了一种在项目公开后使用 Rasch 测量模型对不同测试形式进行等同的方法。此外，本研究检查了 testlet 中的局部项目依赖性，即。即，基于特定刺激（例如阅读段落）和适当的评分方法的一组测试项目。在等同测试形式之前，在具有匹配响应格式的测序项目中观察到局部项目依赖性（LID）。先前的研究表明，管理 LID 的一种方法是将依赖项作为子测试进行评分。然而，本研究的结果表明，将依赖项评分作为子测试或小测试，即。 e.多分评分不一定适用于排序项目。开发了一个特殊程序来对替代测试形式和锚定测试中的排序项目进行评分，以减少 LID。比较精确评分方法和成对评分方法，本研究的结果表明，成对评分会减少一组测序项目中LID的数量。关于等同测试的替代形式，本研究考察了当测试由testlet组成时，不同的锚项目将如何影响测试等同结果：具体来说，本研究考察了以下两个方面：（1）通道长度如何影响等同结果；（2）通道长度如何影响等同结果。 (2) 问题类型如何影响等同结果。创建了五个锚定测试：包含所有项目的基础锚定测试和四个锚定测试。本研究的研究结果如下： 1．四个锚定测试（锚定测试 2-5）显示与基础锚定测试高度相关。关于通道长度，长通道的锚定测试 3 与基础锚定测试的相关性略高于短通道的锚定测试 2。此外，基于锚定测试 2 和 3 的人员能力大致相同。比较排序差异方面的等同精度，结果表明，使用 Anchor Test 2 进行等同比 Anchor Test 3.2 的结果更加稳定。本研究的结果表明，问题的类型会与考生的能力相互作用。当锚定测试仅包含询问主要思想或对文本的一般理解的项目时，本研究的结果表明，等同/链接的准确性似乎会降低，并且需要检查等同/链接结果，特别是对于能力连续体中间范围的考生。较少的

In the testing context where the current study took place, test items are disclosed to test takers immediately after the administration of each test form. The present study shows one way to equate different test forms after the items are disclosed, using Rasch measurement models. Further, the present study examined the local item dependence in testlets, i. e., a set of test items based on a specific stimulus such as a reading passage, and appropriate scoring methods.Prior to equating test forms, local item dependence (LID) was observed among the sequencing items with a matching response format. Previous research studies suggest that one way to manage LID is to score dependent items as a subtest. However, the results of the present study indicate that scoring dependent items as a subtest or testlet, i. e. scoring polytomously, does not necessarily work in relation to sequencing items. A special program was developed to score the sequencing items in the alternate test forms and the ancho … More r test to reduce LID. Comparing exact and pair-wise scoring methods, the findings of the present study indicate that pair-wise scoring would reduce the amount of LID in a set of sequencing items.Concerning equating alternate forms of a test, the present study examined how different anchor items would affect the results of test equating when a test consists of testlets: specifically, the present study examines the following two aspects: (1) how the passage length would affect equating results; (2) how types of questions would affect equating results. Five anchor tests were created: the Base Anchor Test with all the items and four anchor tests. The findings of the present study are as follows:1. The four anchor tests, Anchor Tests 2-5, show high correlations with the Base Anchor Test. Concerning the passage length, Anchor Test 3 with a long passage had a slightly higher correlation with the base anchor test than Anchor Test 2 with a short passage. Also, the person abilities based on Anchor Tests 2 and 3 are approximately the same. Comparing the accuracy of equating in terms of rank order differences, the results indicate that equating with Anchor Test 2 would lead to the results more stable than Anchor Test 3.2. The finding of the present study suggests that the types of questions would interact with test takers' abilities. When an anchor test consists only of items which ask main ideas or general understanding of text, the findings of the present study suggest that the accuracy of equating/linking seems to be reduced and equating/linking results need to be examined especially for the test takers in the middle range of ability continuum. Less