量尺穩定性評估標準之探究

Investigation of the Criteria for Assessing Equating Consistency

此研究是透過模擬資料進行評估等化一致性之探討，即對於潛在量尺漂流(scale drift)的兩個常用標準：直接等化(direct equating)〔相對於間接等化(indirect equating)〕及循環等化(circle equating)〔相對於同一等化(identity equating)〕之研究。在NEAT設計中，當兩群體之能力差異即平均標準化差異(standardized mean difference， SMD)，在所觀察之定錨題分數(anchor scores)有很大差異時，利用連鎖等百分比等化法(chained equipercentile)和線性等化法(linear method)來做等化結果之比較。使用等百分比等化法，當SMD1時，在直接與間接等化間或循環與同一等化間所產生的差異則變得很大。而使用線性等化法時，即使在SMD很小的情況下，上述兩種等化法間所產生的差異仍很大。

The two commonly used criteria in assessing equating consistency for a potential scale drift-direct equating (versus indirect equating) and equating in a circle (versus identity equating) were studied through simulation. The chained equipercentile and linear methods were used for equating when population difference in ability in terms of the standardized mean difference (SMD) in anchor scores were observed with a wide range of values in the anchor test equating design. For the equipercentile method， when the obtained SMD values were less than 1， the differences between the direct and indirect equatings， or between the circular and identity equatings were small and trivial. When the obtained SMD values were greater than 1， the differences of the conversions between the indirect and direct equatings， or between the circular and identity equatings appeared to be substantially large. For the linear method， the differences between the two conversions appeared to be substantially large when the SMD values were very small.