estimateGLM: Poisson and Negative Binomial regression analysis.

In genomaths/MethylIT: Methylation Analysis Based on Signal Detection

Poisson and Negative Binomial regression analysis.

Description

This function is called internally by countTest function. You would need to call it directly only in very special cases.

Perform Poisson and Negative Binomial regression analysis to compare the counts from different groups, treatment and control x: vector of counts groups: factor labeling the members from each group Evaluated models are 'Poisson', 'Quasipoisson', 'Neg.Binomial.W', and 'Neg.Binomial'

Usage

.estimateGLM(
  x,
  groups,
  baseMV,
  w,
  MVrate,
  test = c("Wald", "LRT"),
  p.value = NULL
)

Arguments

x	Matrix of counts.
groups	Groups information derived from a glmDataSet object.
baseMV	Mean and variance of group counts. If baseMean >= baseVar*MVrate, then the nonlinear fit to 'Poisson' and 'QuasiPoisson' models are performed, otherwise only the nonlinear fit to 'Neg.Binomial' and 'Neg.Binomial with weights' models are performed
w	group weights used in glm procedure
MVrate	Minimum Mean/Variance rate.
test	A character string matching one of 'Wald' or 'LRT'. If test = 'Wald', then the p-value of the Wald test for the coefficient of the independent variable (treatment group) will be reported. If test = 'LRT', then the p-value from a likelihood ratio test given by anova function from stats packages will be the reported p-value for the group comparison when the best fitted model is the negative binomial. As suggested for glm, if best fitted model is Poisson or quasi-Poisson, then the best test is 'Chi-squared' or 'F-test', respectively. So, for the sake of simplicity, the corresponding suitable test will be applied when test = 'LRT'.
p.value	Cut off p-value to reject the null hypothesis

Value

GLM model of the group comparison for the given genomic region

Examples

## Get "RangedGlmDataSet" object
data(ds, package = "MethylIT")

X <- ds$counts[69,]
baseMeanAndVar <- data.frame(baseMean = mean(X),
                             baseVar = var(X))

MethylIT:::.estimateGLM(x = X, groups = ds$colData$condition,
                        baseMV = baseMeanAndVar,
                        w = c(1,1), MVrate = 0.95,
                        test = "LRT")

genomaths/MethylIT documentation built on Feb. 3, 2024, 1:24 a.m.