Validity and reliability of GRS-S in sample of teachers and parents – a Czech Contribution

One of the goals of this study was to evaluate the factor structure of the Czech translation of the Gifted Rating Scales-school form (GRS-S; Pfeiffer &amp; Jarosewich, 2003) and to investigate differences in factor structure between teachers (<i>n</i> = 137) and parents (<i>n</i> = 277, of which 256 were mothers). To this end, ten alternative models were tested, each twice – with either continuous or ordinal responses. Three models were found to exhibit acceptable fit and interpretability – the original six-factor model, a bi-factor model with six specific factors, and a bi-factor model with six uncorrelated specific factors and three correlated general factors. The factor structure was comparable for both parent and teacher raters. High correlations between the factors suggest that teacher ratings might be subject to a halo effect, resulting in difficulty to differentiate between the subscale’s purported areas of interest. Further, the concurrent validity of the GRS-S with tests of intelligence (WJ IE II COG or CFT 20-R) and mathematical achievement (TIM^3-5) was evaluated. Ratings on the GRS-S made by teachers show a closer relationship with both criteria than ratings made by parents, although in both cases the relationship with intelligence test scores is closer. Finally, test-retest reliability over a period of multiple months (median 93 days) of the GRS-S was estimated on a sub-sample of teachers (<i>n</i> = 88) and was concluded to be quite high for all GRS-S subscales (<i>r</i> = .84 - .87).

本研究的目标之一是评估《天赋评定量表-学校版》（Gifted Rating Scales-school form, GRS-S; Pfeiffer & Jarosewich, 2003）捷克语译本的因子结构，并考察教师（n=137）与家长（n=277，其中256位为母亲）之间的因子结构差异。为此，我们共测试了十种备选模型，每种模型均基于连续响应或序级响应两种方式各验证一次。最终有三种模型表现出可接受的拟合度与可解释性：原始六因子模型、包含六个特定因子的双因子模型，以及包含六个不相关特定因子与三个相关一般因子的双因子模型。家长与教师评分者的因子结构具有可比性。因子间的高相关提示教师评分可能存在晕轮效应，导致难以区分各分量表预设的关注领域。此外，本研究还评估了GRS-S与智力测验（WJ IE II COG或CFT 20-R）及数学成就测验（TIM³⁻⁵）的同时效度。教师对GRS-S的评分与两项效标的关联度均高于家长评分，尽管两种评分与智力测验分数的关联度均更为紧密。最后，本研究在教师子样本（n=88）中估算了GRS-S历经数月（中位数为93天）的重测信度，结果显示所有GRS-S分量表的重测信度均较高（r=0.84~0.87）。