A Proportional Hazards Regression Analysis of Employee Turnover Among Nurses in New Zealand
研究用比例风险回归分析改进离职测量，结合就业状态与工作持续时间，发现该方法比传统多元回归更准确，适合研究员工离职的学者参考。
Research on employee turnover has usually relied on employment status (quitters vs. stayers) to index the criterion variable (turnover), and on multiple regression (MR) to model the turnover process. In the present study, which was based on a national sample of nurses in New Zealand (n = 1249), we improved the criterion variable by combining employment status with a continuous measure, namely, job survival duration (length of employment starting from the time of survey). We then applied Cox's (1972) proportional hazards regression (PHR) analysis to the improved criterion, and compared the results with those obtained by applying MR on employment status. The comparisons favored PHR: (1) MR selected an extra predictor of quitting that turned out to be nonsignificant, and (2) the numbers of quitters estimated by PHR tended to be more accurate than those by MR.