dtmvnorm.marginal2: Bivariate marginal density functions from a Truncated Multivariate Normal distribution

Description

This function computes the bivariate marginal density function $f(x_q, x_r)$ from a k-dimensional Truncated Multivariate Normal density function (k>=2). The bivariate marginal density is obtained by integrating out (k-2) dimensions as proposed by Tallis (1961). This function is basically an extraction of the Leppard and Tallis (1989) Fortran code for moments calculation, but extended to the double truncated case.

Usage

dtmvnorm.marginal2(xq, xr, q, r, 
 mean = rep(0, nrow(sigma)), 
 sigma = diag(length(mean)), 
 lower = rep(-Inf, length = length(mean)), 
 upper = rep(Inf, length = length(mean)),
 log = FALSE, pmvnorm.algorithm=GenzBretz())

Arguments

Value $x_q$

Value $x_r$

Index position for $x_q$ within mean vector to calculate the bivariate marginal density for.

Index position for $x_r$ within mean vector to calculate the bivariate marginal density for.

mean

Mean vector, default is rep(0, length = nrow(sigma)).

sigma

Covariance matrix, default is diag(length(mean)).

lower

Vector of lower truncation points, default is rep(-Inf, length = length(mean)).

upper

Vector of upper truncation points, default is rep( Inf, length = length(mean)).

log

Logical; if TRUE, densities d are given as log(d).

pmvnorm.algorithm

Algorithm used for pmvnorm

Details

The bivariate marginal density function $f(x_q, x_r)$ for $x \sim TN(\mu, \Sigma, a, b)$ and $q \ne r$ is defined as $$F_{q,r}(x_q=c_q, x_r=c_r) = \int^{b_1}_{a_1}...\int^{b_{q-1}}_{a_{q-1}}\int^{b_{q+1}}_{a_{q+1}}...\int^{b_{r-1}}_{a_{r-1}}\int^{b_{r+1}}_{a_{r+1}}...\int^{b_{k}}_{a_{k}} \varphi{_{\alpha}}_{\Sigma}(x_s, c_q, c_r) dx_s$$

References

Tallis, G. M. (1961). The moment generating function of the truncated multinormal distribution. Journal of the Royal Statistical Society, Series B, 23, 223--229

Leppard, P. and Tallis, G. M. (1989). Evaluation of the Mean and Covariance of the Truncated Multinormal Applied Statistics, 38, 543--553

Manjunath B G and Wilhelm, S. (2009). Moments Calculation For the Double Truncated Multivariate Normal Density. Working Paper. Available at SSRN: http://ssrn.com/abstract=1472153

Examples

Run this code

# NOT RUN {
  
  lower = c(-0.5, -1, -1)
  upper = c( 2.2,  2,  2)
  
  mean  = c(0,0,0)
  sigma = matrix(c(2.0, -0.6,  0.7, 
                  -0.6,  1.0, -0.2, 
                   0.7, -0.2,  1.0), 3, 3)
  
  # generate random samples from untruncated and truncated distribution
  Y = rmvnorm(10000, mean=mean, sigma=sigma)
  X = rtmvnorm(500,  mean=mean, sigma=sigma, lower=lower, upper=upper, 
      algorithm="gibbs")
    
  # compute bivariate marginal density of x1 and x2
  xq <- seq(lower[1], upper[1], by=0.1)
  xr <- seq(lower[2], upper[2], by=0.1)
  
  grid <- matrix(NA, length(xq), length(xr))
  for (i in 1:length(xq))
  {
    for (j in 1:length(xr))
    {
      grid[i,j] = dtmvnorm.marginal2(xq=xq[i], xr=xr[j], 
        q=1, r=2, sigma=sigma, lower=lower, upper=upper)
    }
  }
  
  plot(Y[,1], Y[,2], xlim=c(-4, 4), ylim=c(-4, 4), 
     main=expression("bivariate marginal density ("*x[1]*","*x[2]*")"), 
     xlab=expression(x[1]), ylab=expression(x[2]), col="gray80")
  points(X[,1], X[,2], col="black")
  
  lines(x=c(lower[1], upper[1], upper[1], lower[1], lower[1]), 
        y=c(lower[2],lower[2],upper[2],upper[2],lower[2]), 
        lty=2, col="red")
  contour(xq, xr, grid, add=TRUE, nlevels = 8, col="red", lwd=2)
  
  # scatterplot matrices for untruncated and truncated points
  require(lattice)
  splom(Y)
  splom(X)
# }

Run the code above in your browser using DataCamp Workspace