produced an average total time in system of 16.0831 hours, well under the steady-state mean of 20. Why is our result so much lower than the 20?

(a) Maybe it’s just because of random f uctuation, as we’ve discussed, and we made only one replication that might have just bounced low. Instead, make 50 replications to dampen the randomness, and investigate. (Throughout this exercise you might want to select Run . Run Control . Batch Run (No Animation) to turn off the animation and speed things up.) Since you’re making more than one replication, the Average column in the Category Overview report for Total Time is the average of 50 independent and identically distributed average (within-run) total times, the f rst of which is our 16.0831, and then there are 49 more. The “Half Width” column you’ll see is the “6” part of a 95% conf dence interval on the expected average total time in system, found by taking the standard deviation of the 50 within-run average total times, dividing by the square root of 50, and multiplying by the student t critical value for a 95% conf dence interval with 49 degrees of freedom (this t critical value is 2.010, as you can conf rm by consulting your favorite t table; remember, you want to divide that other 5% equally in the two tails, so you want 2.5% out in each tail). So is it just random f uctuation that caused us to be unlucky and have our f rst replication bounce low, down to 16.0831?

(b) Or maybe it’s because our 160-hour run is not long enough. Lengthen it to 16,000 hours (100 times as long) and go back to just one replication. The Half Width you see here is again that of a 95% conf dence interval, but it’s formed in a different way, discussed in Section 7.2.3. So is it run length?

(c) To see if your result from part (

b) might be subject to random f uctuation as well (HINT: it is, but not as much as in the original short run), replicate the 16,000-hour run in part (

b) 50 times.

(d) Or maybe it’s because of the way we initialized our simulation. Like most simulations, Model