The current situation C and two tentative improvements I 1,2 are assessed by simulation. Some trial runs show that I 1,2 are likely to reduce performance indicator x relative to C, but there is no clear "winner". Confidence from the trial runs of any improvement beating the current situation is less than 90%. Just before the deadline for writing the final report, long simu- lations are carried out for each model, with the following results: C I1I1 IndicatorAverageHalf Width x x I2I2 x How do you report your findings? What do you recommend?
We may safely assume that different simulation runs are independent (at worst). So we can tabulate the differences: C-I 1 Av HW Stdev=HW/1.96 A/S Conf(>0) C-I 2 Z | | | | | | | | | | | | | | | | | | | | | | Report: Both solutions seem to improve current situation (plm. 90% confidence). No clear winner. Implementation depends on costs I 2 -I 1
The current situation C and nine tentative improvements I 1..9 are assessed by simulation. Trial runs show quite erratic behaviors w.r.t. performance indicator x relative to C. In a desparate attempt, just before the deadline for writing the final report, long simulations are carried out for each model, with the following results: C I1I1 IndicatorAverageHW x x I2I2 x What do you do? I3I3 x I4I4 x I5I5 x I6I6 x I7I7 x I8I8 x I9I9 x