metagen (version 1.0)

formulaL: Regression coefficients: formulaL

Description

Calculate pivotal quantities for the regression coefficients using the method: formulaL form the dissertation.

Usage

formulaL(y, d, h, g, x)

Arguments

y

k-vector of responses.

d

k-vector of heteroscedasticity.

h

scalar of heterogeneity.

g

p-vector of some p-variate Gaussian draw.

x

design k-p-matrix.

Value

A p-vector.

Details

Algorithm for calculating a single generalised pivotal quantity for the regression coefficients for given generalised pivotal quantities for the heterogeneity using the univariate version of the pivotal formula.

Examples

bcg   <- bcgVaccineData()
bcg_y <- bcg$logrisk
bcg_d <- bcg$sdiv
bcg_x <- cbind(1,bcg$x)

# When for example using the Mandel-Paule estimate:
bcg_h <- pfunc(y=bcg_y, d=bcg_d, x=bcg_x)(dim(bcg_x)[1] -
  dim(bcg_x)[2])

set.seed(51351) # for reproducibility
random_g <- rnorm(dim(bcg_x)[2])
formulaL(y=bcg_y, d=bcg_d, h=bcg_h, g=random_g, x=bcg_x)

# The function can also be used when planing to perform
# a meta regression with no intercept, and only a singel
# covariate (i.e. dim(x) = 1).  In this case,
# the design matrix can simply be provided by a vector.
set.seed(51351) # for reproducibility
random_g <- rnorm(1)
formulaL(y=bcg_y, d=bcg_d, h=bcg_h, g=random_g, x=bcg$x)

# When performing a meta analysis, provide the function
# with a vector of 1s.
formulaL(y=bcg_y, d=bcg_d, h=bcg_h, g=random_g, x=rep(1,
  length(bcg_y)))