Question

1) how do you decide how long to run a simulation for? (Run length)

2) any two methods for getting a sample for a non terminating solution

3) for a system where no data is available, how do you decide which probability distribution to use? And what do you do if none of the distributions work for you?

4) Why is average utilization a time weighted average? When do we use simple average ?

Answer 1

Answer 1: Markov model which can be analyzed analytically can be used to get preliminary estimates of run lengths by approximating the model of interest.

we can often estimate required run lengths for steady state quantities by calculating the asymptotic variance and the asymptotic bias of the sample mean in the Markov model.

Answer 2: Two methods which can be implemented for getting a sample for a non-terminating solution are:

1.Initialisation bias

2. Batch means

Answer 3: when data is not available knowledge of the process becomes very necessary.

The uniform, triangular, and beta distributions are often used as input probability distribution models when no data is available.

When distribution model fails, similar system approximation (system which behaves similar to our system) and correlation coefficient can be used for modelling.

Answer 4 :

For time and cost statistics, the average (Avg) is the total time or costs divided by the count of resources. Count and utilization statistics are Avg and Tavg.

• Avg is the observation-based average. This is the sum of the value of the statistic, when observed, divided by the number of observations taken.

• Tavg is the time-weighted average. This is the average after weighting each count or utilization based on the amount of time that the observed count or resource usage occurred

The simple average method can work well where:

(a) in each data-set, there is standard data point of interest.
(b) there is very mild fluctuation in data values .