# gldrmCI: Confidence intervals for gldrm coefficients In gldrm: Generalized Linear Density Ratio Models

## Description

Calculates a Wald, likelihood ratio, or score confidence interval for a single gldrm coefficient. Also calculates upper or lower confidence bounds. Wald confidence intervals and bounds are calculated from the standard errors which are available from the gldrm model fit. For likelihood ratio and score intervals and bounds, a bisection search method is used, which takes longer to run.

## Usage

 ```1 2``` ```gldrmCI(gldrmFit, term, test = c("Wald", "LRT", "Score"), level = 0.95, type = c("2-sided", "lb", "ub"), eps = 1e-10, maxiter = 100) ```

## Arguments

 `gldrmFit` A gldrm model fit. Must be an S3 object of class "gldrm", returned from the `gldrm` function. `term` Character string containing the name of the coefficient of interest. The coefficient names are the names of the beta component of the fitted model object. They can also be obtained from the printed model output. Usually the names match the formula syntax, but can be more complicated for categorical variables and interaction terms. `test` Character string for the type confidence interval. Options are "Wald", "LRT" (for likelihood ratio), and "Score". `level` Confidence level of the interval. Should be between zero and one. `type` Character string containing "2-sided" for a two-sided confidence interval, "lb" for a lower bound, or "ub" for an upper bound. `eps` Convergence threshold. Only applies for `test = "LRT"` and `test = "Score"`. Convergence is reached when likelihood ratio p-value is within `eps` of the target p-value, based on the level of the test. For example, a two-sided 95% confidence interval has target p-value of 0.025 for both the upper and lower bounds. A 95% confidence bound has target p-value 0.05. `maxiter` The maximum number of bisection method iterations for likelihood ratio intervals or bounds. For two-sided intervals, `maxiter` iterations are allowed for each bound.

## Value

An S3 object of class 'gldrmCI', which is a list of the following items.

• `term` Coefficient name.

• `test` Type of interval or bound - Wald or likelihood ratio.

• `level` Confidence level.

• `type` Type of interval or bound - two-sided, upper bound, or lower bound.

• `cilo`/`cihi` Upper and lower interval bounds. One one of the two applies for confidence bounds.

• `iterlo`/`iterhi` Number of bisection iterations used. Only applies for likelihood ratio intervals and bounds.

• `pvallo`/`pvalhi` For likelihood ratio intervals and bounds, the p-value at convergence is reported.

• `conv` Indicator for whether the confidence interval limit or bound converged.

## Examples

 ```1 2 3 4 5 6 7 8 9``` ```data(iris, package="datasets") ### Fit gldrm with all variables fit <- gldrm(Sepal.Length ~ Sepal.Width + Petal.Length + Petal.Width + Species, data=iris, link="log") ### Wald 95% confidence interval for Sepal.Width ci <- gldrmCI(fit, "Sepal.Width", test="Wald", level=.95, type="2-sided") ci ```

gldrm documentation built on May 2, 2019, 12:59 p.m.