Rotton J, Frey J
J Pers Soc Psychol. 1985 Nov;49(5):1207-20. doi: 10.1037//0022-3514.49.5.1207.
Archival data covering a 2-year period were obtained from three sources in order to assess relations among ozone levels, nine measures of meteorological conditions, day of the week, holidays, seasonal trends, family disturbances, and assaults against persons. Confirming results obtained in laboratory studies, more family disturbances were recorded when ozone levels were high than when they were low. Two-stage regression analyses indicated that disturbances and assaults against persons were also positively correlated with daily temperatures and negatively correlated with wind speed and levels of humidity. Further, distributed lag (Box-Jenkins) analyses indicated that high temperatures and low winds preceded violent episodes, which occurred more often on dry than humid days. In addition to hypothesized relations, it was also found that assaults follow complaints about family disturbances, which suggests that the latter could be used to predict and lessen physical violence. It was concluded that atmospheric conditions and violent episodes are not only correlated but also appear to be linked in a causal fashion. This conclusion, however, was qualified by a discussion of the limitations of archival data and concomitant time-series analysis.
为了评估臭氧水平、九种气象条件指标、星期几、节假日、季节趋势、家庭纠纷以及人身攻击之间的关系，我们从三个来源获取了涵盖两年时间的档案数据。与实验室研究结果一致，臭氧水平高时记录到的家庭纠纷比低时更多。两阶段回归分析表明，家庭纠纷和人身攻击也与每日气温呈正相关，与风速和湿度水平呈负相关。此外，分布滞后（博克斯 - 詹金斯）分析表明，高温和低风速先于暴力事件发生，暴力事件在干燥天气比潮湿天气更频繁发生。除了假设的关系外，还发现人身攻击发生在关于家庭纠纷的投诉之后，这表明家庭纠纷可用于预测和减少身体暴力。研究得出结论，大气条件与暴力事件不仅相关，而且似乎存在因果联系。然而，通过对档案数据局限性和伴随的时间序列分析的讨论，这一结论受到了限制。
