I’d like to generate random numbers with specific skewness, kurtosis, mean and standard deviation in Excel. Is there an easy way to do this? Does anyone happen to have a script handy?
in general, the moments do not determine the distribution uniquely. even if you know all of the (infinitely many) moments, let alone only the first 4 moments, that doesn’t specify the distribution. so basically you can have many different distributions that have the same mean, variance, skeweness, and kurtosis - your question is way too general i think to have an answer
Pseudorandom, dude. Pseudorandom. As long as you’re using an Excel/VBA routine there’s no way you can effectively generate random numbers, regardless of the distribution.
To some extent, you can simulate kurtosis with an inverse t distribution with few degrees of freedom. Not sure how to get you skew, though.
Try using a mixture of two normal distributions: http://en.wikipedia.org/wiki/Mixture_model You might have to derive parameters that fit your criteria. Then you can simulate two normal distributions independently and mix them yourself using the appropriate weight.
=RANDBETWEEN(m,n)
For some reason I can’t read Maratikus’s link, so not sure if I’m repeating anything here. A common general technique for moment matching is the Edgeworth expansion. A while ago I found an Excel implementation on the web; a bit of googling might track it down for you.
software: crystal ball ($$$) or riskamp
dukatu2 Wrote: ------------------------------------------------------- > =RANDBETWEEN(m,n) 1) you are making an assumption that the distribution is uniform, it needs not be 2) m,n only parametrize first 2 moments, what about skewness and kurtosis
DarienHacker Wrote: ------------------------------------------------------- > A common general technique for moment matching is > the Edgeworth expansion. A while ago I found an > Excel implementation on the web; a bit of googling > might track it down for you. thats a great suggestion, but again, you need to make an assumption about what the distribution is. you can expand the edgeworth series for the normal and match the first 4 moments, or you can match the first 4 terms of the edgeworth expansion for lognormal, exponential, whatever
for a random number in general, ask Mobius Striptease: “what is 1+1?”
Or ask Jeffries Analyst about their bonus.
The.Unit.Root Wrote: ------------------------------------------------------- > software: crystal ball ($$$) or riskamp just tried riskamp, it’s really good. thanks. i found some freeware: http://www.mathtools.net/Excel/Simulation/index.html
nuppal Wrote: ------------------------------------------------------- > Or ask Jeffries Analyst about their bonus. Jefferies (spelled incorrectly) bashing on AF. Love it. Who says the 2007 attitude is dead?
NakedPuts Wrote: ------------------------------------------------------- > nuppal Wrote: > -------------------------------------------------- > ----- > > Or ask Jeffries Analyst about their bonus. > > Jefferies (spelled incorrectly) bashing on AF. > Love it. Who says the 2007 attitude is dead? This place is turning into WSO anyway…
dukatu2 Wrote: ------------------------------------------------------- > =RANDBETWEEN(m,n) ^ this is how our accountants set the yearly budget.