0

In R, when I use rmultinom() I get errors when the size parameter is very big (e.g., 6227020800). Since I would like to study the long term behavior of a random experiment in cases where the size parameter of my multinomial distribution gets really big and very quick, is there a way to get around the limits for the size parameter?

Are there packages that allow a greater size for multinomial distribution generators?

SecretAgentMan
  • 2,856
  • 7
  • 21
  • 41
GGG
  • 101
  • If you want to generate *N* samples and get an error, could you instead generate *N* / 2 samples twice (different rng seeds) and then concatenate the results? – SecretAgentMan Feb 07 '20 at 22:41
  • Not sure I understand what you mean by 'concatenate', however I am afraid dividing by 2 would lead to very little advantage. The kind of growth I have in mind is like size = 2^(2^n)... with a size this big, rmultinom overflows after three or four iterates... I am generating a sequence of multinomials having time dependent size gorwing very fast and time dependent probability parameter, to be more specific. – GGG Feb 09 '20 at 22:48

0 Answers0