Sackett P R, Ostgaard D J
Industrial Relations Center, University of Minnesota, Minneapolis 55455.
J Appl Psychol. 1994 Oct;79(5):680-4. doi: 10.1037/0021-9010.79.5.680.
Correcting validity coefficients for range restriction requires the applicant pool predictor standard deviation (SD). As this is frequently unknown, some researchers use national norm SDs as estimates of the applicant pool SD. To test the proposition that job-specific applicant pools are markedly more homogeneous than broad samples of applicants for many jobs, job-specific applicant pool SDs on the Wonderlic Personnel Test for 80 jobs were compared with a large multijob applicant sample. For jobs at other than the lowest level of complexity, job-specific applicant pool SDs average 10% lower than the broad norm group SDs. Ninety percent of the job-specific applicant pool SDs lie within 20% of the norm group SD, suggesting that reducing a norm group SD by 20% provides a conservative estimate of the applicant pool SD for use in range restriction corrections for other than low-complexity jobs.
校正范围限制的效度系数需要申请人库预测指标的标准差（SD）。由于这一标准差常常未知，一些研究人员使用全国常模标准差作为申请人库标准差的估计值。为了检验特定工作的申请人库比许多工作的广泛申请人样本明显更具同质性这一命题，将80项工作的旺德利克人员测试中特定工作的申请人库标准差与一个大型多工作申请人样本进行了比较。对于复杂度并非最低水平的工作，特定工作的申请人库标准差平均比广泛常模组标准差低10%。90%的特定工作申请人库标准差落在常模组标准差的20%范围内，这表明将常模组标准差降低20%可为除低复杂度工作外的其他工作范围限制校正提供保守的申请人库标准差估计值。
