求助：关于IRT软件

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

请教各位牛人，目前分析IRT的软件有conquest，winsteps，Bilog，Multilog，BilogMG，Parsale等，哪位牛人能告诉我这些软件适用于处理什么样的数据？特别是Bilog，我有些迷糊，敬请赐教，谢谢！

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：IRT软件 IRT winsteps conquest MULTILOG 求助 软件 IRT

补充一下：还有关于Parscale，谁知道这个软件可以同时估计多少题目呢？我这里目前有600多道题目，请问可以同时估计吗？

我这里有在网站上找的一些资料，但是都是偏软件性的，我希望各位达人能提供实践性的经验分享，再次感谢！
Item Response Theory           - 项目反应理论软件包
项目反应理论 (Item Response Theory，称简IRT) 包括一组分析软件包: BILOG-MG, MULTILOG, PARSCALE, and TESTFACT。这些软件作为题目分析、题库建设以及分数估计等方面的重要工具在各个领域被广泛应用。

BILOG-MG
by Michele Zimowski, Eiji Muraki, Robert Mislevy & Darrell Bock
        Graphical user interface
        Efficient analysis of binary items including multiple choice or short-answer items scored right, wrong, omitted, or not-presented
        Capable of large scale production analysis, and handling of multiple groups
        Performs item analysis and scoring of any number of subtests or subscales
        Non-equivalent groups equating
        Vertical equating of test forms
        Differential item functioning (DIF)
        Detection and correction for parameter trends over time (DRIFT)
        Calibration and Scoring of tests in two-stage testing procedures
        Estimation of latent ability or proficiency distributions
        Provision for items inserted in tests to estimate item statistics, but not included in calculation of examinee scores ("variant items")
        Item fit statistics, theoretical and empirical reliability
        Information curves and reliabilities for putative test forms
        Presentation quality IRT graphics, can be imported in Word, Access, etc.
        Detailed online HELP documentation includes description of interface, syntax, and examples.
MULTILOG
by David Thissen, Wen-Hung Chen & Darrell Bock
        Easy to use graphical user interface
        One, two and three-parameter logistic models
        Samejima's model for graded responses
        Bock's model for nominal (non-ordered) responses
        Steinberg's model for multiple-choice items
        Handling of multiple-alternative items, such as multiple-choice tests or Likert-type attitude questionnaires
        Scoring of items with multiple alternatives
        Differential item functioning (DIF)
        Handling of data from several populations simultaneously
        Analysis of mixtures of items types
        Testing of item parameters across groups
        Handling of equality constraints and fixed parameters
        Presentation quality IRT graphics, can be imported in Word, Access, etc.
        Detailed online HELP documentation includes description of interface, syntax, and examples.
PARSCALE
by Eiji Muraki & Darrell Bock
        The flexibility and the wealth of information provided by this program have kept it in regular use by researchers around the world
        One, two, and three-parameter logistic models
        Samejima's model for graded responses
        Master's partial credit model
        Generalized partial credit model
        Analysis of rating scale items such as open-ended essay questions
        Analysis of multiple-choice items
        Differential item functioning (DIF)
        Analysis of mixtures of item types
        Rater's-effect analysis
        Multiple-group polytomous item response models
        Presentation quality IRT graphics, can be imported in Word, Access, etc.
        Detailed online HELP documentation includes syntax and examples.
TESTFACT
by R. Wood, D. Wilson, R. Gibbons, S. Schilling, E. Muraki & D. Bock
        Marginal maximum likelihood (MML) exploratory factor analysis and classical item analysis of binary data
        Computes tetrachoric correlations, principal factor solution, classical item descriptive statistics, fractile tables and plots
        Handles up to 10 factors using numerical quadrature: up to 5 for non-adaptive and up to 10 for adaptive quadrature
        Handles up to 15 factors using Monte Carlo integration techniques
        Varimax (orthogonal) and PROMAX (oblique) rotation of factor loadings
        Handles an important form of confirmatory factor analysis known as "bifactor" analysis: Factor pattern consists of one main factor plus group factors
        Simulation of responses to items based on user specified parameters
        Correction for guessing and not-reached items
        Allows imposition of constraints on item parameter estimates
        Handles omitted and not-presented items
        Detailed online HELP documentation includes syntax and annotated examples.
转自：http://www.pomine.com/show_software.aspx?pName=IRT

论坛太火了，已经沉到大海里去了

DDDDDDDDDDD
有兴趣

网络上倒是有试用版的不知道哪位有正式版的。

论坛太火了，已经沉到大海里去了
本文来自: 人大经济论坛 详细出处参考：http://www.pinggu.org/bbs/viewth ... amp;from^^uid=1328552

自我解答：经过几天的查阅学习，了解到conquest、winsteps多适用于单参模型，parscale比较厉害，可以适用于二元计分的单参、双参、三参；也适用与多级评分的PCM、GPCM模型。不过conquest还可以进行多向度模型的分析。至于bilog系列还没有涉及。。。

2# krong09

理论上估计应该没问题（因为项目参数估计，是1道试题1道试题估计的），如不能估计可能是软件的限制。不过这要自己写命令估计（fortran语言），这是该类软件的缺点（要了解每个命令的基本设置以及含义，该类软件使用的前提就是了解IRT），如果你不懂IRT基本原理，比较麻烦，比如D常数的设置是1还是1.7。就是写出命令，也可能不能运行，........。

关于具体使用什么模型以及软件，要靠检验，看数据适合什么模型。不要一味强调用单参或多参。至于使用数据的维度，确实目前商业化的只有ConQuest（单参数模型），其他关于使用多维资料的软件还没有见到，有的只是个人研究编制的一些程序。

conquest 我们最多一次估计了300个题目