KrzSubspaceBootstrap {evolqg}

Quasi-Bayesian Krzanowski subspace comparison

Description

Calculates the usual Krzanowski subspace comparison using a posterior samples for a set of phenotypic covariance matrices. Then, this observed comparison is contrasted to the subspace comparison across a permutation of the original data. Residuals, which are used to calculate the observed P-matrices, are shuffled across groups. This process is repeated, creating a null distribution of subspace comparisons under the hypothesis that all P-matrices come from the same population. This method is a modification on the fully Bayesian method proposed in Aguirre et. al 2013 and improved in Morrisey et al 2019.

Usage

KrzSubspaceBootstrap(x, rep = 1, MCMCsamples = 1000, parallel = FALSE)

Arguments

x

list of linear models from which P-matrices should be calculated

rep

number of bootstrap samples to be made

MCMCsamples

number of MCMCsamples for each P-matrix posterior distribution.

parallel

if TRUE computations are done in parallel. Some foreach backend must be registered, like doParallel or doMC.

Value

A list with the observed and randomized eigenvalue distributions for the posterior Krz Subspace comparisons.

References

Aguirre, J. D., E. Hine, K. McGuigan, and M. W. Blows. 2013. “Comparing G: multivariate analysis of genetic variation in multiple populations.” Heredity 112 (February): 21–29.

Morrissey, Michael B., Sandra Hangartner, and Keyne Monro. 2019. “A Note on Simulating Null Distributions for G Matrix Comparisons.” Evolution; International Journal of Organic Evolution 73 (12): 2512–17.

See Also

KrzSubspaceDataFrame, PlotKrzSubspace

Examples

library(plyr)
data(ratones)

model_formula = paste("cbind(", paste(names(ratones)[13:20], collapse = ", "), ") ~ SEX")
lm_models = dlply(ratones, .(LIN), function(df) lm(as.formula(model_formula), data = df))
krz_comparsion = KrzSubspaceBootstrap(lm_models, rep = 100, MCMCsamples = 1000)
krz_df = KrzSubspaceDataFrame(krz_comparsion)
PlotKrzSubspace(krz_df)