draw8edcg: simulates random draws for a numcat distribution

In rebastaba: Handling Bayesian networks

Description

(dn) This function simulates a categoric distribution. Similar to the multinomial distribution but the size parameter is one and it is considered as a univariate distribution. Values are the numbers of drawn [lev] categories. If necessary this function could be replaced by some C code... Possible conditional distribution are introduced if !is.null(parents) using their order and unex convention. See the description provided in the function help8ltype for numcat.

Usage

draw8edcg(X, proba, lev=NULL, parents=NULL, parentslength=NULL)

Arguments

X	already generated parents. If no one, must be a matrix nx0 to indicate the number of draws to perform.
proba	Matrix [p,k] of the probabilities of the multinomials. When there are no parents, p=1 because the distribution is the same for all the draws. When there are parents, it is the product of the numbers of categories of the parents. By the way, the parents must be of nature categ.
lev	When null a distribution is drawn with values from 1 to ncol(proba). When numeric transformed as.character(round(lev)). When character, a categoric distributions is drawn with the categories given by lev, a character vector. (If not null the length of lev must be equal to the column number of the proba matrix.)
parents	When null no parents. If not, the parent names as variables names if any. Be careful that their order is significant for the interpretation of the proba matrix.
parentslength	Only when there are parents. Then The vector of lengths of levels of the parents. This cannot be deduced from the simulated values because some of the levels can be missing.

Value

a vector [n] of the draws is returned (it is a factor).

Examples

 rebastaba3k("RESET"); # For R checking
 X <- matrix(NA, 12, 0);
 proba <- matrix(1:5, 1);
 draw8edcg(X, proba);
 UU <- sample(LETTERS[1:4], 20, TRUE);
 X <- as.data.frame(UU);
 proba <- matrix(1:12, 4);
 draw8edcg(X, proba);
 draw8edcg(X, proba, lev=letters[1:12]);

rebastaba documentation built on May 2, 2019, 5:24 p.m.