covariance_etc: Covariance and Standard Deviation MLE
In lcmix: Layered and chained mixture models

Description Usage Arguments Details Value Author(s) See Also Examples

Calcuate the maximum likelihood estimators of covariance and standard deviation.

1 2	mleCov(x, mean=colMeans(as.matrix(x)), simplify=TRUE) mleSd(x, mean=colMeans(as.matrix(x)))

`x`	a numeric vector, matrix, or data frame.
`mean`	the mean, or vector of means, with respect to which covariance should be calculated.
`simplify`	logical; if `TRUE`, the return value will be reduced to the simplest possible form, for example, a single-element matrix will be reduced to a scalar.

The standard R functions cov, var, and sd return unbiased estimators of (co)variance and standard deviation, which are not equal to the maximum likelihood estimator (MLE). The functions given here return MLEs instead. For example, if x is a vector consisting of elements x_1, …, x_N with mean mu, then the MLE of the variance is

sigmasq.hat = (1/N) sum_n (x_n - mu)^2

while the unbiased estimator is

sigmasq.hat = {1/(N-1)} sum_n (x_n - mu)^2.

mleVar is an alias for mleCov.

For mleCov, the estimated variance if x is a vector, or the esitmated covariance matrix of the columns of x if x is a matrix or data frame. If simplify is FALSE, then a matrix will always be returned even if x is a vector.

For mleSd, the estimated standard deviation if x is a vector, or the vector of estimated column standard deviations if x is a matrix or a data frame.

Daniel Dvorkin

thetahat.norm, thetahat.mvnorm for weighted parameter estimation; cov, var, sd in package stats.

set.seed(123)

x <- rnorm(10)
mleCov(x)
# [1] 0.8187336
var(x)
# [1] 0.909704
mleSd(x)
# [1] 0.904839
sd(x)
# [1] 0.953784

x <- matrix(rnorm(30), ncol=3)
mleCov(x)
#            [,1]       [,2]       [,3]
# [1,]  0.9698367 -0.4933797  0.1336627
# [2,] -0.4933797  0.7797652 -0.1931557
# [3,]  0.1336627 -0.1931557  0.2502430
cov(x)
#            [,1]       [,2]       [,3]
# [1,]  1.0775964 -0.5481997  0.1485141
# [2,] -0.5481997  0.8664058 -0.2146175
# [3,]  0.1485141 -0.2146175  0.2780478
mleSd(x)
# [1] 0.9848029 0.8830432 0.5002430
apply(x, 2, sd) # sd(x) is deprecated
# [1] 1.0380734 0.9308092 0.5273024