Monte carlo integration

[Zakaria Zoundi]

Hello! Could Anyone explain me how to perfore monte carlo integration on sata? Any useful link Will be Well appreciated. Thanks

[Carlo Lazzaro]

Zakana:
have you taken a look at -simulate- and related entry in Stata 13.1 .pdf manual?

[Carlo Lazzaro]

Zakana:
-search halton sequence- seems to provide some interesting entries.

[Maarten Buis]

Monte Carlo integration is a very large theme. Some general references are:

There was a special issue in the Stata Journal on Monte Carlo integration using Halton sequences, which you can access here: http://www.stata-journal.com/sj6-2.html

In Stata 14 StataCorp extended the possibilities by adding Markov Chain Monte Carlo as a method of Monte Carlo integration, see: http://www.stata.com/manuals14/bayes.pdf

However, you may get more focusses responses if you give us more details, for example by telling us more about why you want to do Monte Carlo integration

[Zakaria Zoundi]

Thanks all for your prompt reply. really appreciate it. Carlo I am suing stata 12, and so far I havent seen carlo integration in " simulate ". should upgrade to 13 and check.

Maarten, the thing is im very beginner with poor knowledge in programming. I am trying follow paper dealing with a simulation of the impact of development aid on import in less advanced countries. I was looking for a faster and simple way to do Carlo integration. but seems bit hard. Will take time go smoothly.

[Maarten Buis]

simulate and Halton sequences exist in Stata 12, so for those you don't need to upgrade. For MCMC you need to upgrade to Stata 14, but given your background and what you want to do, that is probably overkill.

There are other good reasons for upgrading, but if this is the only one that matters to you, then you are probably better of sticking to Stata 12. That is one of the nice things about the permanent licences that StataCorp gives you: you can upgrade whenever you want, but if you are not going to use the new facilities then nobody is going to force you and you can continue to use a good product that does what you need.

[Carlo Lazzaro]

Zakaria:
you're right: -simulate- (in Stata 12 .pdf manual, too) is only the first step to get some details concerning how Monte Carlo simulation can be programmed in Stata.
As far as Monte Carlo integration is concerned, please consider Maarten's references and sound advice.

[Zakaria Zoundi]

Thanks all so much for these tremendous advice . I'm going smoothly and everything sounds nice so far. A little confusion. What's the assumption behind the number of draws? I repeat as many as possible till the mean of the final outcome approaches the true value? . in some papers 100 times were used, some others more ( 10.000, 100.000 etc) is the number of draws correlated with the probality distribution? Or it's just matter trials. Thanks

[Zakaria Zoundi]

Thanks again, Marteens Actually my concern was about implementing MC integration in sata 12. As the simulate command does only simulation , if I'm not wrong. But I'm reading some other ways to do it.

[Maarten Buis]

simulate is a tool that could potentially be useful, also for MC integration. Whether that is the case for your problem depends on the finer details of your problem.

[Carlo Lazzaro]

Zakaria.
there's no general guidance about the correct number of MC iterations As it is often the case, it depends on the customary rules of your research field. In the economic evaluation of health care programmes, 1000 or 10,000 MC iterations are usually enough to test the robustness of the base case findings via probabilistic sensitivity analysis. This number my well increase under the bayesian framework: it is not infrequent to read about 100,000 iterations of Monte Carlo Markov Chain, but a part of them is then discarded for taking burn-in period into account.

[Zakaria Zoundi]

Ah, I see. thanks so much Maarten and Carlo.