multivariance.timing: estimate of the computation time
In multivariance: Measuring Multivariate Dependence Using Distance Multivariance

Description Usage Arguments Details Examples

View source: R/multivariance-functions.R

Estimates the computation time. This is relative rough. First run with determine.parameters = TRUE (which takes a while). Then use the computed parameters to determine the computation time/or sample size.

multivariance.timing(
  N = NULL,
  n,
  sectime = NULL,
  coef.cdm = 15.2,
  coef.prod = 2.1,
  coef.sum = 1.05,
  determine.parameters = FALSE
)

`N`	number of samples. If `NULL` and `sectime` is given, then `N` is computed.
`n`	number of variables
`sectime`	desired computation time in seconds. If `NULL` then the required computation time is computed.
`coef.cdm`	computation time parameter for the doubly centered distance matrices
`coef.prod`	computation time parameter for matrix products
`coef.sum`	computation time parameter for matrix sums
`determine.parameters`	if `TRUE` then the parameters for the current computer are determined. This might take a while (3 loops to N=1000).

When detecting the parameters, the median of the computation times is used.

Ns = (1:100)*10
ns = 1:100
fulltime = outer(Ns,ns,FUN = function(N,n) multivariance.timing(N,n))
contour(Ns,ns,fulltime,xlab = "N",ylab = "n",
 main = "computation time of multivariance in secs",
 sub = "using default parameters -
 use 'determine.parameters = TRUE' to compute machine specific values")

# Run to determine the parameters of your system:
# multivariance.timing(determine.parameters = TRUE)