EBayes: This function computes empirical Bayes estimates of effect...

In CenterForStatistics-UGent/EBayesH: Emperical Bayes with heterogeneous data

Description

EBayes returns empirical Bayes estimates of effect sizes and their variances.

Usage

EBayes(Z, Y, df = 7, breaks = 120)

Arguments

Z	is a matrix with columns containing the z-values for each predictor from the different studies. the dimension of Z should be "number of predictors by number of studies"
Y	is a list containing the binary response variables for each study. Each element of the list should be of length equal to the number of obsersation in each study. The order of the z-values in the columns of Z must be maintained. That is if the z-values of study k are in column k of Z, then response values of study k should be the kth element of the list Y.
df	is the degrees of freedom of the spline fit used to non-parametrically estimate the marginal density (see Efron (2009)). The default value is df=7. This will work in most cases.
breaks	are the number of bins used in the spline fit (see Efron (2009)). The default is 120 and will also work in most cases if the number of predictors is not below 1000.

Details

Details are found in Mbah et al. (2018). With multiple high-dimensional datasets, this function computes empirical Bayes estimates of each predictor's overall effect size beta which is the ratio between the predictor's mean and its between studies variance. It also computes the predictor variances across studies (heterogeneity). The predictor effect sizes can then be used in a linear prediction rule.

Value

EB_beta	a vector of empirical Bayes estimates of predictor effects
EB_tauSq	a vector of variances across studies for each predictor.

Examples

res <- EBayes(Z,Y)
head(res)

CenterForStatistics-UGent/EBayesH documentation built on May 14, 2019, 6:06 a.m.