Rodgers J L, McGue M
Department of Psychology, University of Oklahoma, Norman 73019.
Behav Genet. 1994 May;24(3):259-62. doi: 10.1007/BF01067192.
DeFries and Fulker's (Behav. Genet. 15, 467-473, 1985) regression procedure (DF analysis) to estimate c2 and h2 was originally applied to selected twin data. Since then, DF analysis has been applied more broadly in unselected data and with multiple (nontwin) kinship levels. Theoretical work based on the matrix algebra of variance-covariance matrices has shown that estimates of c2 and h2 are unbiased in selected two-group settings. In this article, a simple proof is presented supporting the validity of DF analysis in broader settings. We use scalar algebra to show that parameter estimates of h2 and c2 are unbiased in unselected settings with multiple (more than two) kinship levels. Caveats are offered, and other DF analysis problems are identified.
德弗里斯和富尔克（《行为遗传学》第15卷，第467 - 473页，1985年）用于估计c2和h2的回归程序（DF分析）最初应用于选定的双胞胎数据。从那时起，DF分析已更广泛地应用于未选定的数据以及具有多个（非双胞胎）亲属关系水平的数据。基于方差 - 协方差矩阵的矩阵代数的理论工作表明，在选定的两组情况下，c2和h2的估计是无偏的。在本文中，给出了一个简单的证明，支持DF分析在更广泛情况下的有效性。我们使用标量代数表明，在具有多个（多于两个）亲属关系水平的未选定情况下，h2和c2的参数估计是无偏的。同时给出了注意事项，并识别了其他DF分析问题。
