logiDx: Diagnostics for logistic regression

In logisticDx: Diagnostic tests for logistic regression models

Description

Returns standard diagnostic measures for a logistic regression model by covariate pattern

Usage

  logiDx(x, round = FALSE, roundTo = 3)

Arguments

x	A model of class glm
round	If round=TRUE, digits will be ronded to roundTo decimal places
roundTo	No. decimal places to which to round digits

Value

A data.table. There is one row per covariate pattern with at least one observation. These are sorted by dBhat (see below).

The initial columns give all combinations of the predictor variables with at least one observation.

Subsequent columns are labelled as follows:

obs	Number of observations with this covariate pattern
prob	Probability of this covariate pattern
yhat	Number of observations of y=1, predicted by the model
y	Actual number of observations of y=1 from the data
lev	Leverage, the diagonal of the hat matrix used to generate the model; a measure of influence of this covariate pattern
devR	Deviance residual, calculated by covariate pattern; a measure of influence of this covariate pattern
PeR	Pearson residual, calculated by covariate pattern; a measure of influence of this covariate pattern. Given by: obs^0.5 (prob/1-prob)^0.5
sPeR	Standardized Pearson residual calculated by covariate pattern; a measure of influence of this covariate pattern. Given by: PeR.(1-lev)^0.5
dBhat	Change in Bhat, the standardized difference between the original maximum likelihood estimates B and that estimates with this covariate pattern excluded
dXsq	Change in chi-square, decrease in the value of Pearson chi-square statistic with this covariate pattern excluded. Given by: sPeR^2
dDev	Change in deviance D with this covariate pattern excluded. Given by: d^2/(1-lev)

Note

Values for the statistics are calculated by covariate pattern. Different values may be obtained if calculated for each individual obervation (i.e. row in data frame).

Generally, the values calculated by covariate pattern are preferred, particularly where no. observations are >5.

See Also

plotLogiDx

Examples

d1 <- genLogiDt(model=FALSE)
f1 <- stats::glm(y ~ I(x5^2)*x1 -1, family=binomial("logit"), data=d1)
logiDx(f1)

logisticDx documentation built on May 2, 2019, 6:30 p.m.