Variation of the initiation time of vector control measures. Panels A and D show simulations that reproduce the original 2003 and 2008/2009 outbreaks. Panels B,C,E and F show simulated 2003 and 2008/2009 like outbreaks where the onset of control was varied by +/-2 weeks from the original start day. In both cases, variation of the control start day had a significant effect on the outbreak curve and final case numbers, confirming previous studies, such as . In the plots, black dots are observed outbreak data, solid black lines are best stochastic realisations, dotted lines are the medians of 60 simulations, dark gray shaded areas are interquartile ranges, light gray shaded areas are 95% confidence ranges. Note that the predicted cases shown in Panels A-F are the median case numbers of 60 stochastic realizations. Often there is a considerable spread, e.g. in Panel E the median (middle dotted line, which lies almost on the x-axis) predicted number of cases is 32 but in rare cases, case numbers can exceed 800.