電腦適性測驗概述

背景

電腦適性測驗(computerized adaptive test,CAT)為了提供比傳統測驗更準確以及更有效率的測量

題庫發展

執行步驟

initial step

test step

  1. 以目前受測者的答題情況估計受測者的能力
  2. 如果終止條件未滿足,
    • 下一題按照選題方式,自題庫中選擇符合條件的題目
    • 更新受測者的答題情況
    • 回到步驟1
  3. 如果終止條件滿足,終止施測,計算受測者能力

能力估計方式

選題方式

列舉幾種文獻上常見的選題方式,各方式有其優缺點。目前最常見的是The maximum Fisher information (MFI) criterion。

  1. The maximum Fisher information (MFI) criterion
  2. The so-called bOpt criterion, also sometimes referred to as Urry’s rule
  3. The maximum likelihood weighted information (MLWI) criterion
  4. The maximum posterior weighted information (MPWI) criterion
  5. The Kullback-Leibler (KL) divergency criterion
  6. The posterior Kullback-Leibler (KLP) criterion
  7. The maximum expected information (MEI) criterion
  8. The minimum expected posterior variance (MEPV) criterion
  9. The so-called thOpt procedure
  10. The progressive method
  11. The random selection among available items: the benchmark selection method for simulation studies

stopping step

有四種終止條件。一般來說一個測驗只會設定一種終止條件。但也可以設定複合式的終止條件,只要條件之一滿足即終止測驗

final step