Quasibinomial detection-count strategy

Quasibinomial detection-count strategy

Quasibinomial detection-count strategy

Value

An R6 class generator.

Details

Fits detected and undetected child-feature counts for each parent feature using glm with a quasibinomial family. The symmetric pseudo-count stabilizes fits under complete separation; it is not equivalent to Firth's bias-reducing penalty.

Super class

prolfqua::StrategyBase -> StrategyBinomial

Public fields

prior_count

symmetric pseudo-count added to both outcomes

Methods

Public methods

• StrategyBinomial$new()

• StrategyBinomial$clone()

Inherited methods

• prolfqua::StrategyBase$contrast_fun()
• prolfqua::StrategyBase$df_residual()
• prolfqua::StrategyBase$isSingular()
• prolfqua::StrategyBase$model_fun()
• prolfqua::StrategyBase$sigma()

---

Method new()

Create a quasibinomial count strategy. sigma() is the Pearson residual scale used by vcov().

Usage

StrategyBinomial$new(
  modelstr,
  prior_count = 0.1,
  model_name = "binomial_nested"
)

Arguments

modelstr

right-hand-side model formula, for example "~ group_"

prior_count

non-negative symmetric pseudo-count

model_name

model identity

---

Method clone()

The objects of this class are cloneable with this method.

Usage

StrategyBinomial$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

dat <- data.frame(
  group_ = factor(rep(c("A", "B"), each = 4)),
  detected = c(1, 2, 1, 3, 4, 5, 3, 5),
  undetected = c(4, 3, 4, 2, 1, 0, 2, 0)
)
strategy <- StrategyBinomial$new("~ group_")
fit <- strategy$model_fun(dat)
coefficients(fit)
#> (Intercept)     group_B 
#>  -0.5937747   2.2264695