pseudo_data_gen: Generate pseudo-data using summary statistics

In uvastatlab/uvastatlab: Miscellaneous functions for the UVA Library StatLab

Generate pseudo-data using summary statistics

Description

This function generates data from summary statistics that in turn have the same summary statistics. This can be useful to replicate analyses for which we only have summary statistics. See reference and example. Code comes from https://github.com/lizlimpocouhasselt/LMM-from-summary-statistics.

Usage

pseudo_data_gen(n = 1, mu, Sigma, tol = 1e-06, empirical = TRUE)

Arguments

n	number of observations to generate
mu	mean
Sigma	covariance matrix
tol	used to confirm Sigma is positive definite
empirical	logical (default is TRUE). Which procedure to use.

Value

Matrix

References

Limpoco, M.A.A., Faes, C. and Hens, N. (2024), Linear Mixed Modeling of Federated Data When Only the Mean, Covariance, and Sample Size Are Available. Statistics in Medicine. https://doi.org/10.1002/sim.10300

Examples

# using the Swiss data that come with R
xbar <- apply(swiss, 2, mean) # mean
s <- cov(swiss)               # covariance

# Now generate psuedo data using these summary stats
pseudo_swiss <- pseudo_data_gen(n = nrow(swiss), mu = xbar, Sigma = s)

# verify summary stats of pseudo data match original summary stats
all.equal(xbar, apply(pseudo_swiss, 2, mean))
# [1] TRUE
all.equal(s, cov(pseudo_swiss))
# [1] TRUE

# Same analysis results
pseudo_swiss <- as.data.frame(pseudo_swiss)
m1 <- lm(Fertility ~ ., data = swiss)
m2 <- lm(Fertility ~ ., data = pseudo_swiss)
all.equal(coef(m1), coef(m2))
# [1] TRUE

summary(m1)$coefficients
summary(m2)$coefficients

uvastatlab/uvastatlab documentation built on June 1, 2025, 10:57 p.m.