[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"Nick Cox" <n.j.cox@durham.ac.uk> |

To |
<statalist@hsphsun2.harvard.edu> |

Subject |
RE: st: RE: Generation of Random Lognomral Distributions |

Date |
Sat, 30 Apr 2005 19:14:14 +0100 |

As said, the recipe is . gen lognormal = exp(<mean> + <sd> * invnorm(uniform())) So you need to loop over that generating a set of variables. Say forval i = 1/100 { local mean = <plug in> local sd = <plug> gen lognormal`i' = exp(`mean' + `sd' * invnorm())) } You refer to a uniform distribution. The precise definitions to be plugged in above would depend on whether that was defined on the income or log income scale. I don't think you can _guarantee_ a value < 300,000. Nick n.j.cox@durham.ac.uk Jose Marin > I apologize for the inconvenience. > > what I am trying to do is generate random distributions that look like > income distributions at different periods of time. In that case the > distributions have to be lognormal and have a lower bound since > negative income is not feasible, likewise a positive infinity income > is not feasible ether and the distributions have to be different among > themselves. So I want the distributions to be generated randomly where > the numbers in the distribution have a lower limit of zero and an > upper limit of 300,000 and that they have different means and standard > deviations that come from a uniform distribution since the > distributions should have the same chance of being generated. > > This is the problem in its fullest expression. > > Thank you again. Jose > > On 4/30/05, Nick Cox <n.j.cox@durham.ac.uk> wrote: > > It makes sense but it is not precise, > > as you don't say what distribution(s) the > > mean and standard deviation come from. > > > > I could make guesses at what you mean, > > but it would be better if you would just > > state the problem precisely and in one go. Jose Marin > > > > > Thanks Nick, > > > Now what i meant about mean and sd whithin a range is > that I need the > > > distributions to be generated randomly. That is, that the mean and > > > standard deviations be randomly chosen from a range. The > end result is > > > a set of random numbered distribution with randomly > chosen mean and > > > standard deviation. > > > > > > Does this make any sense?? > > > > Nick Cox > > > > > > A random sample from a lognormal can be generated directly by > > > > > > > > . gen lognormal = exp(<mean> + <sd> * invnorm(uniform())) > > > > > > > > Here replace <mean> and <sd> by variables or constants giving > > > > the desired mean(s) and sd(s) of the logged variable. > > > > > > > > I am not clear exactly what you mean by "within a ra[n]ge", > > > > but some variation on this will get what you want. > > > > > > > > (Naturally, this is the recipe used by -rndlogn-. In > > > > the help file for that program, and in the code, > > > > whenever it says variance, it means sd.) > > > > Jose Marin > > > > > > > I have a quick question. > > > > > I am trying to generate a large number of random lognormal > > > > > distributions with mean and standard deviation within a > > > rage. I tried > > > > > using mkbilogn and rndlgn but they generate distributions > > > around the > > > > > same inputted mean and standard deviation. I am a new > Stata user. * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**Follow-Ups**:**st: [stata 9] windowing questions***From:*"George M Hoffman MD" <ghoffman@mcw.edu>

- Prev by Date:
**st: Standard error of prediction minus constant?** - Next by Date:
**st: [stata 9] windowing questions** - Previous by thread:
**Re: st: RE: Generation of Random Lognomral Distributions** - Next by thread:
**st: [stata 9] windowing questions** - Index(es):

© Copyright 1996–2017 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |