Skip to contents

Compute mean and covariance of a matrix

Source: R/meancov.R
meancov.Rd

meancov computes the mean and covariance of a matrix, optionally with weights. This function is used to convert raw data into aggregate form (mean and covariance) for use in aggregate data modeling.

Usage

meancov(m, wt)

Arguments

m

A numeric matrix or data frame containing the observations. Each row represents an individual, and each column represents a time point. For pharmacometric data, columns typically represent concentration measurements at different time points.

wt

Optional vector of weights for each observation. If not provided, all observations are weighted equally. Weights can be used to account for different sample sizes or reliability of different data sources.

Value

A list containing:

  • E: Vector of means for each time point (population typical values)

  • V: Covariance matrix representing the variability between individuals

Details

The function computes:

  • The mean of each column (time point) using colMeans for unweighted data or weighted means for weighted data

  • The covariance matrix using cov.wt with maximum likelihood estimation, which provides unbiased estimates of the population covariance

The maximum likelihood estimation method is used because:

  • It provides unbiased estimates of the covariance matrix

  • It is appropriate for aggregate data modeling where we want to estimate population parameters

  • It handles both balanced and unbalanced designs through optional weights

Key features:

  • Handles missing data automatically through the underlying cov.wt function

  • Provides numerically stable computations

  • Can be used with both raw PK data and simulated data

  • Supports weighted calculations for meta-analysis or combined analysis

Examples

# Load required libraries
library(admr)
library(rxode2)

# Create a matrix of concentration measurements
# 10 subjects measured at 10 time points
m <- matrix(rnorm(100), nrow = 10, ncol = 10)

# Compute unweighted mean and covariance
# Useful for single-study analysis
result <- meancov(m)

# Compute weighted mean and covariance
# Useful for meta-analysis or when combining studies
weights <- runif(10)  # weights could represent study sizes
result_weighted <- meancov(m, weights)