4

I am trying to write R code to generate a nonnormal distribution (i.e., a normal distribution with specific skewness and kurtosis). I would like this distribution to have the following parameters: mean (0), variance (1), skewness (3), and kurtosis (11). Is there a package that can do this, where I simply specify the values?

Sidenote: I've read about Fleishman's Power Method and its coefficient table (found here: https://chandrakant721.wordpress.com/fleishmans-power-method-coefficient-table/), but these particular skewness and kurtosis values are not listed and (I infer) not able to generate proper coefficients. Any help is appreciated.

lmo
  • 37,904
  • 9
  • 56
  • 69
HAL-9000
  • 99
  • 2
  • 9
  • 3
    Did this question not help? http://stackoverflow.com/questions/4807398/how-to-generate-distributions-given-mean-sd-skew-and-kurtosis-in-r?rq=1. Asking for package recommendations is really off-topic. This really seems more like a statistical question about random number rather than anything about programming in R specifically. Maybe ask at [stats.se] instead. – MrFlick Feb 21 '17 at 21:26
  • @MrFlick Thanks for your response. I had seen that previous post, and unless I'm missing something, the answer posted there doesn't illustrate hard-coding skewness and kurtosis values, but rather combining vectors of rnorm distributions. – HAL-9000 Feb 21 '17 at 22:04
  • I also meant to post (above) that I know I can generate rnorm distributions in R and use the rnormmix package, but in this instance I do not want to mix two normal distributions but create a nonnormal distribution outright. – HAL-9000 Feb 21 '17 at 22:05
  • The normal distribution has skewness 0 and kurtosis 3; outside those values, there's *no such thing* as "a normal distribution with specific skewness and kurtosis"; what information is "normal" intended to convey in that expression? Your question is unclear. You'd also need to clarify whether it was the sample skewness and kurtosis or the population values you were trying to specify. ... ctd – Glen_b Feb 22 '17 at 00:54
  • ctd ... If you ask it as a general question (how to do it rather than asking about R code) and appropriately clarify it (an infinity of different distributions have skewness 3 and kurtosis 11 for example), you could then flag to migrate to stats.stackexchange.com (though similar questions have been asked before, so the question may already be answered there). – Glen_b Feb 22 '17 at 00:56
  • The answer from the linked question that referred to using the Pearson distribution does answer this question. It does seem like a duplicate but not off-topic since it is specifically regarding the R code to generate random function. – user3386170 Jan 08 '20 at 14:59

0 Answers0