vbicp.unb: Variational Bayesian approximate mixed-effects inference on...

In kaybrodersen/micp: Bayesian mixed-effects inference on classification performance

Variational Bayesian approximate mixed-effects inference on classification accuracy using the normal-binomial model

Description

Variational Bayesian approximate mixed-effects inference on classification accuracy using the normal-binomial model

Usage

vbicp.unb(ks, ns, verbose = NULL)

Arguments

ks	Vector of successes in each population member.
ns	Vector of attempts in each population member.
verbose	Level of output verbosity. Deprecated parameter.

Details

The return 'effects' represent accuracies in logit space, which has infinite support. In order to obtain, e.g., a posterior-mean estimate of the population accuracy in the conventional [0..1] space, use: 'logitnmean(q$mu.mu, 1/sqrt(q$eta.mu))'

Value

A list of posterior moments:

mu_mu: mean of the posterior population mean effect

eta_mu: precision of the posterior population mean effect

a_lamba: shape parameter of the posterior precision population effect

b_lambda: scale parameter of the posterior precision population effect

mu_rho: vector of means of the posterior subject-specific effects

eta_rho: vector of precisions of posterior subject-specific effects

Author(s)

Kay H. Brodersen, ETH Zurich

References

K.H. Brodersen, J. Daunizeau, C. Mathys, J.R. Chumbley, J.M. Buhmann, & K.E. Stephan (2013). Variational Bayesian mixed-effects inference for classification studies. NeuroImage (2013). doi:10.1016/j.neuroimage.2013.03.008. https://kaybrodersen.github.io/publications/Brodersen_2013_NeuroImage.pdf

Examples

q <- vbicp.unb(c(6, 8, 5), c(10, 10, 10))

kaybrodersen/micp documentation built on April 15, 2022, 2:24 a.m.