I've been digging into the YKS (University Entrance Exam) structure and I'm curious about the exact way raw test scores are transformed into the final scaled scores used for university placement. Specifically, what statistical methods or normalization techniques are applied, and how do they handle variations across different test sessions? Any detailed explanations or references would help me and others trying to demystify the process. How do you usually approach this kind of analysis, and what resources do you recommend for a deeper dive?
Understanding the Scoring Algorithm of YKS: How Are Raw Scores Converted?
👁️ 66 görüntüleme💬 1 cevap❤️ 0 beğeni
1 Cevap
The YKS raw scores are first converted to a session‑specific “scaled” score through a linear transformation that mirrors the equating process used in tests like the U.S. SAT. For each test form ÖSYM publishes the mean and standard deviation of the raw‑score distribution; a Z‑score is calculated ( \(Z = (X‑\mu)/\sigma\) ), then this Z is mapped onto a fixed interval (usually 0–100 for each sub‑test) by adding a constant and multiplying by a factor that keeps the overall range comparable across sessions. Anchor items that appear in multiple administrations are used to anchor the scale, so any difficulty drift or session‑to‑session variance is absorbed before the final scaling step. In practice the formula looks something like \(Scaled = 100 + (Z × 15)\), which yields a mean around 100 and a standard deviation of about 15, just as the SAT’s “scaled score” does after its IRT‑based equating.
If you want to dig deeper, the official ÖSYM “YKS Scoring Guide” (available on their website) spells out the exact constants used each year, and the academic paper “Equating the Turkish University Entrance Exam” (Çelik & Şahin, 2022) walks through the statistical underpinnings. For a hands‑on approach, I usually download the raw‑score PDFs for a given year, compute the session means/SDs in Excel or R, and then replicate the linear scaling to see how a 70‑point raw score translates into the final 120‑point sub‑test score. Comparing this with the SAT’s IRT‑based equating shows the same goal—making scores comparable across different test dates—but the YKS relies on a simpler linear model rather than the full IRT calibration used abroad. This contrast can be a good sanity check when you’re trying to explain the process to peers.