A Monte Carlo Study of Missing Item Methods
通过蒙特卡洛模拟比较了五种处理量表项目缺失数据的方法，发现回归插补在恢复心理测量属性和统计功效方面最有效，而列表删除虽无偏但导致统计功效严重下降。
A Monte Carlo simulation methodology was used to examine the effectiveness of five methods for handling missing data in items that comprise summated scales: listwise deletion, item mean substitution, person mean substitution, random number substitution, and regression imputation. Overall, regression imputation was the most effective method in reproducing psychometric properties of a summated scale and preserving statistical power to detect correlations between the summated scale and other variables. The item mean substitution method yielded biased estimates of the standard deviation of the summated scale. Although list- wise deletion yielded relatively unbiased estimates of the psychometric properties of the summated scale, there were high levels of dispersion in the estimates around the true population values. Furthermore, listwise deletion led to severe reductions in statistical power to detect correlations between the summated scale and other variables.