By water resources engineers; favorite quote: "to report your observations does not qualify for research"-My sensei-

Friday, July 16, 2010

Day 1: Freddy & Monte Carlo

Part 1:



Part 2:


You may download the video using Real Player (free).
 
Hello friends,

First meeting: 5 assistants. We have recorded the presentation without sound. Some of the questions asked during our session are written below.


As you may have noticed, the short explanation shown in the video was not prepared beforehand. This was made in purpose in order to show all that we do not need to spend much time.

Thank you for your time guys. Until the next meeting....

Freddy

Q&A

1. What is a model?
A representation of a system (simple or complex).

2. What is sensitivity analysis?
The investigation of the variability in the outputs of a model, through changes in the inputs within given bounds (uncertainty bounds).

3. What are the steps to carry Monte Carlo simulation?
Step1. Assume the extreme bounds for the input parameters (uncertainty bounds for the inputs) for which you wish to perfom the analysis.
Step2. Assume an independent probability distribution for each input model parameters. In general, we can start assuming that all follow a uniform distribution.
Step3. Using a generator of random numbers, generate a sample for each parameter. In general, the sample should be larger than 1000.
Step4. Each sample is constituted by a set of model parameters. Each set of model parameter values are introduced to your model, and one model output (may be constituted by many outputs) is generated by your mathematical model.
Step5. The model outputs are analyzed through statistical measures (e.g., variance, standard deviation. etc).

4.Monte Carlo and sensitivity analysis?
It is always important to perform sensitivity analysis before a model is applied (e.g., a rainfall-runoff model). If the model is simple, it may be enough to perform a one-at-a-time analysis: changing the values of one parameter at a time, and observing how the outputs change. If the model is complex, we may need to apply more sophisticated methods. Monte Carlo and its variations are a good alternative.

5.References.
Basics:
Wittwer, J.W. 2004. , "Monte Carlo Simulation Basics" From Vertex42.com, June 1, http://vertex42.com/ExcelArticles/mc/MonteCarloSimulation.html (this is the web page I show in the video).
Sobol. 1993. A primer for the Monte Carlo method.
Sensitivity analysis:
Saltelli. 2004. Sensitivity Analysis in Practice - A Guide to Assessing Scientific Models
Software:
Simlab: Sensitivity analysis, sample generations.

No comments:

Post a Comment

Related Posts Plugin for WordPress, Blogger...