scorestatcorr: Score test statistics in generalized linear models, with...
In oyvind-bakke/fwerapprox: FWER Approximations

Description Usage Arguments Value Examples

Computes score test statistics for testing whether each of a large number of coefficients (typically corresponding to genetic markers) in a GLM with canonical link is zero in presence of a smaller number of covariates (typically environmental covariates), and provides estimates of correlations between the test statistics.

1	scorestatcorr(formula, xg, maxorder, family = gaussian, both = FALSE)

`formula`	Model `formula` for the null model, i.e. including environmental covariates.
`xg`	Matrix of genetic markers (one column for each marker).
`maxorder`	Maximal order of which correlations of score test statistics are estimated.
`family`	Family of GLM passed to `glm`.
`both`	If `TRUE`, correlations are given both as a list and as a matrix (see 'Value').

A list containing the following components:

`statistic`	A vector of score test statistics
`corrs`	A list of vectors of correlations of score test statistics. The first component is a vector of first-order correlations, i.e. between neighbouring markers, the second is a vector of second-order correlations, i.e. between markers of distance 2, etc. The number of components is `maxorder`. The list is generated if `maxorder` is greater than 0 and less than the number of columns of `xg` or if `both` is `TRUE`.
`corrmatrix`	Estimated correlation matrix of the score test statistics. The matrix is generated if `maxorder` is equal to the number of columns of `xg` or if `both` is `TRUE`. In the latter case, any correlations between markers of distance greater than `maxorder` will be set to 0. `scorestatcorr` will attempt to load the `Matrix` package and return `corrmatrix` as a sparse banded matrix.

.

# Normal model with three environmental covariates:
result <- scorestatcorr(y_normal ~ sex + activity + agecategory, xg, 2)
result$statistic[1:10]
result$corrs[[1]][1:10]
pvals <- 2*pnorm(-abs(result$statistic)) # p-values for two-sided test
pvals[1:10]

# Logistic model without environmental covariates:
resultl <- scorestatcorr(y_logistic ~ 1, xg, 2, family = binomial)
resultl$statistic[1:10]
resultl$corrs[[1]][1:10]
pvalsl <- 2*pnorm(-abs(resultl$statistic))
pvalsl[1:10]