Assault as a function of time and temperature: A moderator-variable time-series analysis.

The authors hypothesized that relations between temperature and assaults are stronger during evening hours than during other hours of the day and tested this hypothesis by obtaining 3-hr measures of assaults, temperature, and other weather variables for a 2-year interval. The hypothesis was confirmed by autoregressio n analyses that controlled for secular trends, seasonal differences, other weather variables, holidays, and other calendar events. In addition, as predicted by the negative affect escape model, assaults declined after reaching a peak at moderately high temperatures. The inverted Ushaped relation survived tests that controlled for secular trends, seasonality, autocorrelation, outliers, and heteroscedasticity. In addition, consistent with routine activity theory, moderator-variable regression analyses indicated that relations were strongest during evening hours and on weekends.