Xcalibre 在對評估數據進行項目反應理論(IRT)分析的軟體中居於領先地位。Xcalibre 使用4個二分(dichotomous)和5個多分類(polytomous) IRT模型來調整您的檢驗，並可以產生專業匯總分析報告，包括內嵌圖形和表格來幫助數據解釋。Xcalibre 4是目前最綜合的IRT軟體，完全不需要執行DOS命令!
XCALIBRE 使用邊際最大概似法估計項目參數(Item parameter estimation)，此法的優點是可以改進當測試的數目很少時或短缺時。XCALIBRE依然可以估計2-3參數IRT模型，非常適合大量且有希疏性的不完整資料估計。
Xcalibre 4 提供了多個超越Xcalibre 1.1的特色：
1. Superior output: save time by utilizing the rich text file (RTF) format prepared as a formal report with embedded tables and graphics, rather than having to build your own report manually. Graphics include the item response function (IRF), the item information function (IIF), the test information function (TIF), and conditional standard error of measurement (CSEM), and numerous frequency distributions. In addition, you can also overlay the empirical fit line with the IRF.
2. More models: Dichotomous models include the 3PL, 2PL, 1PL, and Rasch models. Five polytomous models (below) are also included, plus you are allowed to mix the models for mixed-format tests.
3. Screening: automatically screen out items with unacceptable classical statistics before performing the IRT item parameter calibration.
4. Custom flagging: customize item flagging rules by specifying acceptable ranges and your own flag labels.
5. Produces spreadsheets: easily import results into test development software because statistics are saved as a comma-separated value (CSV) format, which is also able to be easily manipulated (sorted, highlighted, etc.) in spreadsheet software.
6. More theta estimates: can now be produced for multiple domains as well as the full test.
7. Mastery testing: automatically implement cutscores for criterion-referenced tests.
8. More items: analyze up to 1,200 items at once.
9. Pilot/pretest items: compute classical statistics for pretest (pilot or unscored) items to be used for item evaluation.
10. Multiple runs (batch): utilize a “batch” type of capability, to allow you to run multiple data sets without having to use the graphic user interface for each run.
11. Differential item functioning (DIF): evaluate DIF with the Mantel-Haenszel approach.
- 3-parameter dichotomous model, ideal for multiple choice data
- 2-parameter dichotomous model, ideal for personality data
- 1-parameter dichotomous model, with equivalent item discriminations
- Rasch dichotomous model, scaled to items rather than people
- Rasch rating scale model (RRSM, or RSM)
- Rasch partial credit model (RPCM, or PCM)
- Generalized rating scale model (GRSM)
- Generalized partial credit model (GPCM)
- Samejima's Graded response model (SGRM, or GRM)
Moreover, you are allowed to utilize both a dichotomous and polytomous model at once. For example, if your test has 100 multiple choice questions and 5 open response questions scored with partial credit, you can calibrate all items easily onto a common scale. Because no programming code is involved (point-and-click only), even researchers new to IRT can easily produce such powerful analyses.
- Mantel-Haenszel differential item functioning (DIF)
- Automatically calculate scaled scores for examinees
- Items with insufficient response frequencies now excluded
- Statistics, scores, and test information files are now optional
- Option for comma-delimited or tab-delimited output
- Meaningful labels for domains provides easier to read output
- Skew, Q1, Median, Q3, IQR now computed for all scores
- -2LL provided to compare fit of competing models
- Item control and item status tables provided in report
- Item-person map plots the distribution of b and θ together
- Item parameter and flag table presents the most important
Xcalibre 4 is designed to be user-friendly, with a straightforward graphical user interface that involves no DOS-style command code. All options are specified by radio buttons and check boxes; once you have all of your specifications set, simply hit the "Run" button!
Xcalibre 4 produces top-of-the-line reports designed to make it easier to apply IRT. Instead of oversimplified ASCII-text files, it produces a detailed document with colored headings, embedded figures and tables, and explanations. All statistics are also saved as a CSV file to allow you to work in a spreadsheet and easily upload to test development systems such asFastTEST Web . In addition to the detailed page show below, which is produced for every test item,Xcalibre 4 provides a wide range of test-level output for the overall test as well as each domain (content area):
- Test information function (TIF)
- Conditional standard error of measurement (CSEM)
- Test response function (TRF)
- Grouped frequency distributions of theta estimates
- Coefficient alpha
- Summary statistics of item parameters, raw scores, and theta estimates.
- Overall grouped frequency distributions of item parameters
Xcalibre 4 licenses are determined by the use of the program:
Student research license - Proof of student status required.
Academic - The program will be used at an accredited college or university for research or teaching purposes only.
Government/Non-profit - The program will be used at a governmental or a non-profit organization, or by a school district or college or university for analyzing the data from its testing programs. The use of the program must also be non-profit; that is, if the organization is non-profit but conducts consulting activities for other organizations, the use is commercial.
Commercial - The program will be used for commercial purposes, for example, in the course of consulting in test development and psychometrics.