Partialling Factor Scores Does Not Control Method Variance: A Reply to Podsakoff and Todor
通过数学推导和蒙特卡洛模拟，发现剔除第一主成分的方法会引入负偏差，损害后续分析，且偏差不随样本量增加而减少，因此不推荐使用。
Podsakoff and Todor (1985) proposed partialling the first principal component from observed correlations as a procedure for controlling method variance. Using mathematical derivations and Monte Carlo simulation, we found that this procedure is biased. Partialling out the first principal component introduces a negative bias into the resulting correlations that seriously compromises subsequent analysis. Moreover, the extent of bias is not reduced by increasing sample size; however, it is inversely proportional to the number of variables. Therefore, partialling the first principal component is not recommended. Researchers are encouraged to collect data with multiple methods whenever feasible.