三類測驗型式
以下摘要撰寫計畫過程中整理的測驗形式(包含linear test、CAT、MST)摘要,並簡述優、缺點。
值得紀錄的是
- CAT的優點不只是施測較有效率,同時也可以有較高的測驗準確度。詳細的原因涉及測驗理論,不在此詳述。
- MST在書中乍看不出優點,但在與研究伙伴討論後想到幾個可能的用法在此記錄。
Linear test
- 所有受測者回答所有的題目,無論受測者的能力高低以及題目的難度
- 需要大量的題目才能夠用總分準確地測量受測者能力
- 對於能力在母群當中接近平均值的受測者的測量較準確,對能力接近極端值者的能力測量較不準確
computerized adaptive test (CAT)
- 受測者接受測驗同時估計受測者的能力,利用演算法來決定受測者下一題需回答的題目
- CAT相較於Linear test有兩個優勢
- 通常較有效率(通常需施測較少之題目)
- 測量較準確
multistage test (MST)
- 結合上述兩種測驗形式的特性
- 先測量常規題組(routine test),再根據這一組題目的測量結果選擇後續的測驗題組
- 常規題組可以做為初步篩選,可能可以有幾種用法
- 利用常規題組將受測者分類(如健康人或是病人),再進行相對應合適的測驗內容
- 在多能力測驗中,先測量基礎題目,再視受測者表現選擇不同能力之題目
Table : A comparison of linear, CAT, and MST designs (from Magis et al., 2017, p 4)
Linear Test
Advantages | Disadvantages |
---|---|
Ease of assembly | Full length test |
Ease of administration | Inefficient for measurement |
Least effort for test development (TD) | Inflexible test schedule for test takers |
Prone to test copying |
CAT
Advantages | Disadvantages |
---|---|
Shorter test length | Complicated to implement |
Efficient for measurement | Depends on strong model assumptions |
Flexible test schedule for test takers | Requires a large calibration data set |
Avoids test copying | Greatest effort for TD |
Item exposure more difficult to control | |
Costly to administer via computer | |
Robustness concerns |
MST
Advantages | Disadvantages |
---|---|
Intermediate test length | Depends on model assumptions |
Efficient for measurement | Longer than CAT but shorter than linear test |
Allows test taker item review (within modules) | Item exposure concerns (similar to CAT) |
Easier to implement | Costly to administer via computer (similar to CAT) |
Easier to assemble | |
Moderate effort for TD | |
Flexible test schedule for test takers | |
Reduces test copying |
references
Magis, D., Yan, D., & Von Davier, A. A. (2017). Computerized Adaptive and Multistage Testing with R. Springer International Publishing. https://doi.org/10.1007/978-3-319-69218-0