具性別恆等測量基礎的空間能力性別差異

本研究應用多樣本結構方程模式與性別恆等性檢驗，探討臺灣高中一年級學生在空間能力的性別差異。本研究使用標準化空間能力測驗之常模資料，共計樣本為1,639人，其中男生865人，女生774人；本測驗為「空間視覺」與「空間定位」雙因素結構、四題型、共計18 題選擇題的紙筆式測驗。研究結果顯示，雙因素結構的標準化空間能力測驗具有性別恆等性，亦即具有跨性別的形貌、因素負荷量、截距、測量誤差、因素變異數以及因素共變數恆等；男女的雙因素潛在平均數與觀察平均數均達顯著差異，亦即男女在「空間視覺」的潛在平均數與觀察平均數均達顯著差異，在「空間定位」的潛在平均數與觀察平均數亦均達顯著差異，並且皆為男優於女。綜合而論，臺灣高中一年級男生的空間視覺與空間定位能力均顯著地優於女生。潛在平均數結構分析較觀察平均數分析能夠更客觀且合理地反應出不同群組的水準差異。具恆等性的測驗得分之組間差異，可用潛在平均數與觀察平均數分析，兩者結果將為一致；但若測驗不恆等，則無從得知僅使用觀察平均數進行組間差異比較的後果。本研究結果為標準化空間能力測驗擴增性別比較的效度基礎與推廣應用性，也確認該測驗能提供品質良好且具性別恆等基礎的測驗資料。

關鍵字

臺灣高一學生；多樣本結構方程模式；性別恆等性；空間能力的性別差異；測量恆等性

並列摘要

The study tested gender invariance of the Standardized Spatial Ability Test (SPAT) with multi-group structural equation modeling (MG-SEM) to compare gender differences of the first-year senior high school students in terms of spatial ability latent factor means as well as observed means.The study utilized the norm data of SPAT whose final effect sample size of students is 1,639 (865 boys and 774 girls). The structure of SPAT is two-factor (spatial visualization and spatial orientation). There are 18 multiple-choice items classified into four item types. The results showed that the tests of gender invariance (configuration, factor loadings, item intercepts, error variances, factor variance matrices, and factor covariance matrices) are satisfied, and there are significant differences between genders in their latent factor means and observed means. That is, the latent means and observed means of spatial visualization and spatial orientation were significantly different between genders with the same direction that boys were better than girls in all of the differences. In conclusion, the spatial ability of first-year senior high school boys is better than that of girls in terms of spatial visualization and spatial orientation.The latent mean structure analyses can provide more reliable and reasonable group comparisons than observed mean analyses. These two analyses would provide the same result of difference comparisons when a test is confirmed invariance; however, when a test is not confirmed invariance, there is no way to tell the consequences of using only observed mean analyses for group difference comparisons. The results of the study increase the validity of gender comparisons of SPAT and its applications, and provide evidences of gender-invariant and quality measurement for SPAT.