Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

adaptivetest

bởi woodstocksoftware

Personal Development
339Lượt tải do clawskills.sh công bố
0Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Adaptive testing engine with IRT/CAT, AI question generation, and personalized learning recommendations.

Mô tả

Kỹ năng này làm gì

Covers the design and implementation of computerized adaptive tests (CAT) using Item Response Theory. Includes IRT models (1PL, 2PL, 3PL), item selection algorithms, ability estimation methods, and stopping rules. Applies to K-12 placement, certification, diagnostic, and progress monitoring assessments.

Adaptive tests achieve comparable precision to fixed-length tests with roughly half the items, shortening test time without sacrificing measurement accuracy.

Khi nào nên dùng

  • Building a math placement test that adjusts question difficulty in real time
  • Calibrating an item bank for a certification or licensure exam
  • Estimating student reading level with fewer questions than a fixed-length test
  • Setting content balance constraints across multiple subject areas
  • Choosing between fixed-length and standard error stopping rules for a diagnostic assessment

Quy trình ví dụ

ĐẦU VÀO

User asks: design a K-12 math placement CAT with 300 items, SE < 0.35 target, max 25 questions, and content balance across algebra, geometry, and statistics

AGENT
  1. 1

    Define test specifications: purpose (placement), ability range (-2 to +2), max length 25 items, SE target 0.35

  2. 2

    Specify item bank distribution: 300 items spread across difficulty range with 40% algebra, 30% geometry, 30% statistics

  3. 3

    Select algorithms: Randomesque (top 5) for item selection, EAP for ability estimation, combined stopping rule

  4. 4

    Outline simulation plan: run 1000 sessions with known abilities to check average test length and item exposure rates

  5. 5

    Define output format: ability estimate plus placement recommendation

ĐẦU RA

A complete CAT design spec with item bank requirements, algorithm choices, and stopping rules that places 95% of students within 0.5 grade levels of true ability in an average of 18 questions

Cài đặt

clawhub install woodstocksoftware/adaptivetest

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết