BayesianCalculateMatrix function

Calculate Covariance Matrix from a linear model fitted with lm() using different estimators

Calculates covariance matrix using the maximum likelihood estimator, the maximum a posteriori (MAP) estimator under a regularized Wishart prior, and if the sample is large enough can give samples from the posterior and the median posterior estimator.

BayesianCalculateMatrix(linear.m, samples = NULL, ..., nu = NULL, S_0 = NULL)

Arguments

• linear.m: Linear model adjusted for original data
• samples: number os samples to be generated from the posterior. Requires sample size to be at least as large as the number of dimensions
• ...: additional arguments, currently ignored
• nu: degrees of freedom in prior distribution, defaults to the number of traits (this can be a too strong prior)
• S_0: cross product matrix of the prior. Default is to use the observed variances and zero covariance

Returns

Estimated covariance matrices and posterior samples

Examples

data(iris)
iris.lm = lm(as.matrix(iris[,1:4])~iris[,5])
matrices <- BayesianCalculateMatrix(iris.lm, nu = 0.1, samples = 100)

References

Murphy, K. P. (2012). Machine learning: a probabilistic perspective. MIT press.

Schafer, J., e Strimmer, K. (2005). A shrinkage approach to large-scale covariance matrix estimation and implications for functional genomics. Statistical applications in genetics and molecular biology, 4(1).

Author(s)

Diogo Melo, Fabio Machado