AGENDA Analysis for terminating models ANALYSIS I Analysis for terminating models… Analysis for non-terminating models ANALYSIS FOR TERMINATING MODELS Replication analysis Production simulation runs Statistical analysis of the simulation run results Economic analysis of statistical analysis results REPLICATION ANALYSIS OF TERMINATING MODELS REPLICATION ANALYSIS OF TERMINATING MODELS Select an initial number of replications… Calculate summary statistics… Calculate the level of precision… If the precision is less than the desired precision calculate new number of replications… Run the new required number of replications… Repeat the process until the level of precision meets the desired level of precision… SELECT AN INITIAL NUMBER OF REPLICATIONS 10 CALCULATING SUMMARY STATISTICS Mean of replication means Standard deviation of replication means Half width confidence interval or standard error CALCULATE RELATIVE PRECISION Formula COMPARE CALCULATED AND DESIRED LEVELS OF PRECISION If calculated level of precision is less than desired level of precision Last number of replications is sufficient If calculated level of precision is greater than desired level of precision Last number of replications is insufficient How many replications to run in next iteration? NUMBER OF REPLICATIONS IS INSUFFICENT FOR RELATIVE PRECISON When relative precision exceeds the desired relative precision Manipulate equation to calculate new required number of replications PRODUCTION SIMULATION RUNS OF TERMINATING MODELS The replication analysis must be performed for each individual alternative Some alternatives will require few replications than others All of the alternatives must be run at the highest number of replications required by any single alternative Must run at least 10 replications for any alternatives CALCULATE SUMMARY STATISTICS EXAMPLE RELATIVE PRECISION FIXED NUMBER OF REPLICATIONS NEXT Determine the number of replications required for all the other alternatives Rerun all the alternatives at the highest number Recheck all of the alternative values to verify that the mathematical 0.09 precision is achieved If they do not, you must repeat the process