glm.predict (version 4.3-2)

dc.polr: predicted values and discrete change

Description

The function calculates the predicted values and the difference of two cases with the confidence interval. It can be used for a polr model.

Usage

# S3 method for polr
dc(model, values = NULL, sim.count = 1000, conf.int = 0.95, 
  sigma = NULL, set.seed = NULL, values1 = NULL, values2 = NULL,
  type = c("any", "simulation", "bootstrap"), summary = TRUE)

Value

The output is a matrix have in the first column the predicted values, in the second column the lower value of the confidence interval and in the third column the upper value of the confidence interval.

Arguments

model

the model-Object generated with polr()

values

the values of case 1 and 2 as vector in the order how they appear in the summary(model) Estimate. Values is if values1 and values2 are specified after each other in the same vector. Either values or values1 and values2 have to be specified.

sim.count

OPTIONAL numbers of simulations to be done by the function. default: 1000

conf.int

OPTIONAL the confidence interval used by the function. default: 0.95

sigma

OPTIONAL the variance-covariance matrix, can be changed when having for example robust or clustered vcov. default: vcov(model)

set.seed

OPTIONAL set a seed for the random number generator

values1

the values of case 1 as vector in the order how they appear in the summary(model) Estimate. Has to be defined if values is not defined.

values2

the values of case 2 as vector in the order how they appear in the summary(model) Estimate. Has to be defined if values is not defined.

type

OPTIONAL choose between simulation and bootstrap, "any" chooses between those two according to the number of cases (bootstrap if n < 1000)

summary

OPTIONAL if mean/quantiles should be return or all simulated values (default: TRUE)

Author

Benjamin Schlegel,

Details

The function makes a simulation for the two cases and compares them to each other.

Examples

if (FALSE) {
data = MASS::survey
data$Smoke = ordered(data$Smoke, levels = c("Never", "Occas", "Regul", "Heavy"))
model1 = polr(Smoke ~ Height + Pulse, data=data)
summary(model1)
dc(model1, values1 = c(150,mean(MASS::survey$Pulse,na.rm=TRUE)),
  values2 = c(151,mean(MASS::survey$Pulse,na.rm=TRUE)))
# all differences are significant as the confidence intervals do not include 0
}