Facets 性能評價分析軟體 
Facets 是基於 Windows的軟體，在Rasch-Model的很多應用過程中幫助，特別是在性能評價的區域內並且配對比較。建議您先熟悉較簡單的WINSTEPS才開始使用Facets。

Facets旨在處理從簡單到非常艱難的一維多面Rasch測量的應用。它從涉及不同方面組合的複雜數據構建度量，例如考生，項目，任務，評委以及進一步的測量和結構方面。它旨在在一次分析中處理不同格式的項目組合。項目類型包括二分法，最多255個類別的評級量表，泊松計數和伯努利試驗。多個不同的測量模型可以包含在同一分析中，包括配對比較，排序，李克特和其他評定量表，部分信用和二分法，所有這些都涉及1到255個方面。例如，雙面人物項目多項選擇題可以包含在與四面人物評價者任務項目評級相同的分析中。度量也可以單獨或按組均值固定（錨定），便於在測試會話中進行等同和鏈接。提供所有措施的質量控制適合性評估。識別出意外的數據點。可以測量偏差，差異項目功能和相互作用。可以實施加權方案。一次分析中可以包括多達400萬個單獨的元素，例如考生，評估者等。典型應用程序的數據丟失率超過90％。在數據矩陣的每個單元中允許多個觀察。Facets已成功用於構建語言測試，醫務人員績效，患者表現，公開演講，個人和團隊的體育表現，以及消費者的喜好。

Aids to Conceptualization
Facets is designed for number-crunching, but it does include some visual aids. The “rulers” show how the measures lay out up the variable. This enables judge leniencies, examinee abilities, and item easiness to be compared with
each other, and with the rating scale. Also displayed are numerical bar charts of the counts, scores measures and fit statistics for each element (person, etc.) of each facet (examinees, etc.)

Output Tables & Plots
Output Tables and Bias analyses can beproduced at estimation time. This can be done in “silent” batch mode. Reports can be customized in an interactive phase. For large analyses, it can be easier to look at a single output table, than to scroll through voluminous standard output

Output Files Output files in several formats giving complete information about measures and responses.. Additional DIF/Bias analyses can be performed. Simulated data can be produced. Facets specification and data files can be converted into Winsteps control and data files.

Unexpected Data Points Identified Garbage in - Garbage out! Quality-control is central to Facets philosophy. Valid results require valid data. Data entry errors, erratic judges, mis-specified rating scales, deficient experimental designs .... Facets identifies suspected problems for investigation, but always produces results.

Bias analysis - Differential Item Function
Interactions of any facet with any other measurement or demographic facet can be investigated and automatically plotted with Excel. Comprehensive Help file with examples. Free downloadable Manual in PDF format. Student/evaluation versions downloadable free from www.winsteps.com

Constructs Measures from Awkward Data
Facets is for tough applications of measurement construction from complex ordinal data involving heterogeneous combinations of examinees, items,tasks, judges, etc. along with further
measurement and structural facets. Over 90% missing data - no problem. Up to 1 million examinees etc. Data in Excel or text files. Facets is used for analyzing educational, medical and
psychological assessments, also for staff evaluations, public speaking, artistic and sports competitions, consumer preferences and more ....

One or More Item Formats in One Analysis
Facets is designed to handle flexibly combinations of items of different formats in one analysis. Item types include dichotomies, rating scales with up to 255 categories, partial credit
items, Poisson counts and Bernoulli trials. Full reporting of item structures includes category probability curves, expected and empirical item characteristic curves (ICCs) and information functions.

Multiple Measurement Models in an Analysis
Multiple different measurement models in the same analysis, including paired-comparisons,rank- order, rating scales, partial credit and
dichotomizations involving from 1 to 255 facets. Unjudged multiple-choice items can be combined with judge-awarded essay ratings (weighted, if desired) to construct linear measures. Measures adjusted for judge leniency, task difficulty, etc.

Measures, Quality-Control Fit Statistics
Linear measures, standard errors, quality-control fit statistics (Infit, Outfit), point-biserials, and item discrimination estimates are reported. Also summary reliability and agreement statistics.
Measures can be fixed (anchored) individually or by group mean, for equating and linking across test sessions. The “observed average” shows the average rating for an element (person, item,
judge, etc.). The “fair average” adjusts that for the context - leveling the playing-field.