Beruflich Dokumente
Kultur Dokumente
Aknowledgement
Wayne L. Winston, Microsoft Excel Data Analysis
Overview
Part I
Questions answered with the help of MCS History Typical simulations
Part II: Simulation examples Part III: Advantages of MCS over deterministic
analysis
Challenges
We are constantly faced with uncertainty, ambiguity,
and variability.
Risk analysis is part of every decision we make.
situations that present uncertainty and play them out thousands of times on a computer.
cars to order?
What is the probability that a new products cash
possible outcomes of your decisions and assess the impact of risk, allowing for better decision making under uncertainty.
simulations were performed to estimate the probability that the chain reaction needed for the atom bomb would work successfully.
The Monte Carlo method was coined then by the
physicists John von Neumann, Stanislaw Ulam and Nicholas Metropolis, while they were working on this and other nuclear weapon projects (Manhattan Project) in the Los Alamos National Laboratory.
It was named in homage to the Monte Carlo Casino, a
famous casino in the Monaco resort Monte Carlo where Ulam's uncle would often gamble away his money.
MCS Use
General Motors (GM), Procter and Gamble (P&G),
and Eli Lilly use simulation to estimate both the average return and the riskiness of new products.
MCS Use: GM
Forecast net income for the corporation Predict structural costs and purchasing costs
portfolios.
By definition, Value at Risk at security level p for a
Pr(X<VaR_p(X))=p
clients retirement.
project
MCS Applications
Physical Sciences Engineering Computational Biology Applied Statistics Games Design and visuals Finance and business (Actuarial Science) Telecommunications Mathematics
Part II
Well now discuss how Monte Carlo simulation
=RAND() function
When you enter the formula =RAND() in a cell, you
get a number that is equally likely to assume any value between 0 and 1.
Get a number less than or equal to 0.25 around 25% of
the time
Get a number that is at least 0.9 around 10% of the
time
discrete r.v.:
10,000 20,000
40,000 60,000
Less than 0.10 Greater than or equal to 0.10 and less than 0.45
Greater than or equal to 0.45 and less than 0.75 Greater than or equal to 0.75
0.1
0.45 0.75
20,000
40,000 60,000
2
0.076074298 3 0.364201634 4 0.698116365
10,000
20,000 40,000
col_index_num, range_lookup )
1)=60,000
TRUE=1, FALSE=0
lookup is TRUE, it uses the largest value that is less than or equal to lookup value.
then calculate the fraction of time each demand appears in the simulation, well get a table similar to the following:
DEMAND DEMAND 10,000 FRACTION OF TIME 0.10250 10,000 20,000 40,000 PROBABILI TY 0.10 0.35 0.30
20,000
40,000 60,000
0.35500
0.29250 0.25000
60,000
0.25
for a normal r.v. with a mean =40,000 and standard deviation =10,000
What is a normal random variable?
Let us first define the standard normal random
variable.
the zero.
Standard Normal Distribution Table is a table that
shows probability that a standard normal random variable Z is less than a number z:
(z)=Pr(Z<z)
A standard normal r.v. Z is a r.v. with =0 and =1
for a normal r.v. with a mean =40,000 and standard deviation =10,000
The formula NORMINV(RAND(), , ) will generate
VLOOKUP(rand, lookup, 2)
The number of unites sold is
cost
If we produce more cards than are demanded, the
Disposal cost in 9:
- 12,321.19 48,346.89 73,622.44 10,000 25000 25000 25000 25000 25000 20,000 50000 50000 50000 50000 50000 40,000 16000 100000 16000 100000 100000 60,000 -60000 66000 66000 150000 -18000
Cards to Produce? (cont.) Enter 1-1000 on the left corresponding to our 1,000
trials
Data menu.
Click on any blank cell (e.g. I14) as the column
input cell and choose production quantity (cell B1) as the row input cell.
We calculate the average simulated profit for each
production quantity
We calculate the standard deviation of simulated
expected profit
However, it also appear to have a large standard
deviation (risk)
expected profits drop by about 22%, but the risk drops almost 73%.
Therefore, if we are extremely risk averse,
std.dev. of zero cards because if we produce 10,000 cards we will always sell all of them and have none left over.
will fall?
This interval is called the 95% confidence interval
Problems
1
A GMC dealer believes that demand for 2005 Envoys will normally be distributed with a mean of 200 and standard deviation of 30. His cost of receiving an Envoy is $25,000, and he sells an Envoy for $40,000. Half of all leftover Envoys can be sold for $30,000. His is considering ordering 200, 220, 240, 260, 280, and 300 Envoys. How many should he order?
Problems (cont.)
2
A small supermarket is trying to determine how many copies of Newsweek magazine they should order each week. They believe their demand for Newsweek is governed by the following discrete random variable
DEMAND 15 20 25 30 PROBABILITY 0.10 0.20 0.30 0.25
35
0.15
Problems (cont.)
2
The supermarket pays $1.00 for each copy of Newsweek and sells each copy for $1.95. They can return each unsold copy of Newsweek for $0.50. How many copies of Newsweek should the store order to maximize its profit?
Advantages of MCS
MCS provides a number of advantages over deterministic, or single-point estimate analysis:
Probabilistic Results Graphical Results
Sensitivity Analysis
Scenario Analysis
Correlation of Inputs
Probabilistic Results
Results show not only what could happen, but how
Graphical Results
Because of the data a Monte Carlo simulation
generates, its easy to create graphs of different outcomes and their chances of occurrence.
This is important for communicating findings to
other stakeholders.
Sensitivity Analysis
With just a few cases, deterministic analysis makes
Scenario Analysis
In deterministic models, its very difficult to model
different combinations of values for different inputs to see the effects of truly different scenarios.
Using Monte Carlo simulation, analysts can see
exactly which inputs had which values together when certain outcomes occurred.
This is invaluable for pursuing further analysis.
Correlation of Inputs
In Monte Carlo simulation, its possible to model
Its important for accuracy to represent how, in reality, when some factors go up, others go up or down accordingly.
References
Wayne L. Winston, Microsoft Excel Data Analysis
http://www.palisade.com/risk/monte_carlo_simulatio n.asp