Classical Item Analysis

Classical item analysis describes each test item by its difficulty, discrimination and correlation with the rest of the test, and shows how reliability would change if the item were removed.

고전 문항분석은 각 문항의 난이도, 변별도, 나머지 문항과의 상관을 기술하고 그 문항을 뺐을 때 신뢰도가 어떻게 바뀌는지 보여 줍니다.

Which questions on my 30-item exam are too easy, too hard or fail to separate strong from weak students?

30문항 시험에서 너무 쉽거나 어렵거나 상위·하위 학생을 가려내지 못하는 문항은 어느 것인가요?

Item difficulty is the mean score relative to the maximum, which for right-or-wrong items is the proportion answering correctly. Discrimination is assessed with the corrected item–total correlation and with the difference between the top and bottom 27% of total scorers, and alpha-if-deleted flags items that weaken internal consistency.

When to use Classical Item Analysis

• Reviewing a test or questionnaire before final use
• Identifying items that do not discriminate or may be miskeyed
• Reporting item statistics for a classroom or certification exam

When to use something else

• You need item parameters that do not depend on the sample, where item response theory models are appropriate
• Only an overall reliability coefficient is needed, where Cronbach's alpha or KR-20 suffices

Assumptions and requirements

• Items measure a single construct
• Items are scored in the same direction
• Total score is a reasonable ranking of ability
• Each item's maximum possible score is known; otherwise the observed maximum is used for difficulty

What the analysis reports

• Item difficulty (p value for binary items)
• Item–total and corrected item–rest correlations
• Upper–lower 27% discrimination index
• Alpha if item deleted
• Overall Cronbach's alpha

How to report it

Across the [value] items (N = [value]), mean item difficulty was [value], corrected item–total correlations ranged from [value] to [value], discrimination indices ranged from [value] to [value], and Cronbach's α was [value].

Related methods

• Cronbach's Alpha Reliability Analysis
• Kuder-Richardson 20 (KR-20)
• Item Response Theory (IRT)

Run Classical Item Analysis on Quorant