Compared　with　parametric　analysis　method,　nonparametric　method　is　more　robust　and　widely　used.　In　order　to　make　different　safety　policy　of　various　kinds　of　time　and　reduce　traffic　crashes,　this　paper　wisely　and　reasonably　uses　nonparametric　tests　models,　such　as　Mann-Whitney,　and　Shapiro-Wilk,　to　analyze　traffic　crash　date.　After　testing　of　normality,　this　paper　discusses　timely　distributions　of　traffic　crashes　-hourly,　monthly　and　yearly　distributions　of　traffic　crash　data　by　nonparametric　tests.　So　some　right　models　for　suitable　timely　distribution　are　given.　Finally,　analysis　of　related　crashes　data　in　America　and　China　is　carried　out.　©　2009　ASCE.