概化理论方差分量变异量估计方法

Estimation Procedure of the Variability of Estimated Variance Components for Generalizability Theory

方差分量变异量估计是概化理论分析的要害。Traditional方法、Bootstrap方法、Jack-knife方法和MCMC方法可以用于估计概化理论方差分量变异量。根据方法取向的不同,可以将四种概化理论方差分量变异量估计方法分为两大类:再抽样方法和近似估计方法。基于再抽样方法和近似估计方法的优势和不足,参考一些相关文献,将Bootstrap方法和MCMC方法相结合估计概化理论方差分量变异量,其可靠性和精确性将有可能得到进一步的提高。

The variability of estimated variance components is the ＂ Achilles heel＂ of generalizability theory. To estimate the variability of estimated variance components for generalizability theory, traditional, bootstrap,jackknife and Markov Chain Monte Carlo （MCMC） procedures can be used. In terms of the difference of these procedures, we can divide them into two categories, that is resampling method and approximate estimation method. According to the advantage and disadvantage of resampling method and approximate estimation method, some researches shows an integrated method may be developed to improve the reliability and accurancy of estimating the varia- bility of estimated variance components for generalizability theory,that is merging bootstrap procedure with MCMC procedure.