| 1 | #' Generate a sample of (X,Y) of size n with default values |
| 2 | #' @param n sample size |
| 3 | #' @param p number of covariates |
| 4 | #' @param m size of the response |
| 5 | #' @param k number of clusters |
| 6 | #' @return list with X and Y |
| 7 | #' @export |
| 8 | generateXYdefault = function(n, p, m, k) |
| 9 | { |
| 10 | meanX = rep(0, p) |
| 11 | covX = diag(p) |
| 12 | covY = array(dim=c(m,m,k)) |
| 13 | for(r in 1:k) |
| 14 | covY[,,r] = diag(m) |
| 15 | pi = rep(1./k,k) |
| 16 | #initialize beta to a random number of non-zero random value |
| 17 | beta = array(0, dim=c(p,m,k)) |
| 18 | for (j in 1:p) |
| 19 | { |
| 20 | nonZeroCount = sample(1:m, 1) |
| 21 | beta[j,1:nonZeroCount,] = matrix(runif(nonZeroCount*k), ncol=k) |
| 22 | } |
| 23 | |
| 24 | sample_IO = generateXY(meanX, covX, covY, pi, beta, n) |
| 25 | return (list(X=sample_IO$X,Y=sample_IO$Y)) |
| 26 | } |
| 27 | |
| 28 | #' Initialize the parameters in a basic way (zero for the conditional mean, uniform for weights, |
| 29 | #' identity for covariance matrices, and uniformly distributed for the clustering) |
| 30 | #' @param n sample size |
| 31 | #' @param p number of covariates |
| 32 | #' @param m size of the response |
| 33 | #' @param k number of clusters |
| 34 | #' @return list with phiInit, rhoInit,piInit,gamInit |
| 35 | #' @export |
| 36 | basicInitParameters = function(n,p,m,k) |
| 37 | { |
| 38 | phiInit = array(0, dim=c(p,m,k)) |
| 39 | |
| 40 | piInit = (1./k)*rep(1,k) |
| 41 | |
| 42 | rhoInit = array(dim=c(m,m,k)) |
| 43 | for (i in 1:k) |
| 44 | rhoInit[,,i] = diag(m) |
| 45 | |
| 46 | gamInit = 0.1 * matrix(1, nrow=n, ncol=k) |
| 47 | R = sample(1:k, n, replace=TRUE) |
| 48 | for (i in 1:n) |
| 49 | gamInit[i,R[i]] = 0.9 |
| 50 | gamInit = gamInit/sum(gamInit[1,]) |
| 51 | |
| 52 | return (list("phiInit" = phiInit, "rhoInit" = rhoInit, "piInit" = piInit, "gamInit" = gamInit)) |
| 53 | } |