Page 1 of 1

Parameters with triagular distribution

PostPosted: Tue Sep 12, 2017 5:29 am
by lauraab
I am trying to run a RP model in which one of the parameters follows a triangular distribution, is it possible to include in this model??

Re: Parameters with triagular distribution

PostPosted: Thu Sep 14, 2017 1:34 pm
by Michiel Bliemer
Short answer: No, it is not possible to use triangular distributions in Ngene, currently only uniform and normal.

Long answer: We are working to include exponential and lognormal as well, but triangular is more difficult since this specific distribution cannot be created based on a single standard distribution. For example, (log)normal distributions can be generated from a single random variable X1 that is uniformly distributed, and we rely on this property to do all the smart draws in Ngene (Halton, Sobol, Gaussian quadrature, etc). A triangular distribution requires two random variables X1 and X2 that are uniformly distributed, making it much more difficult to efficiently do smart draws. Since triangular distributions are only used by a relatively small number of researchers, it is not on the top our list to include in Ngene.

Michiel