GDAtools (version 1.5)

translate.logit: Translate logit regression coefficients into percentages

Description

Performs a logit regression and then computes the effects of covariates expressed in percentages (through two methods: 'pure' effects and 'experimental' effects; see Deauvieau, 2010)

Usage

translate.logit(formula,data,nit=0)

Arguments

formula

an object of class formula (or one that can be coerced to that class): a symbolic description of the model to be fitted. Every variables have to be factors.

data

a data frame containing the variables in the model

nit

number of bootstrap iterations for confidence interval computation. Default is 0, i.e. no confidence interval is computed.

Value

The function returns a list:

reg

An object of class glm or nnet (depending on the number of factors of the dependent variable)

summary

The results of summary function applied to reg element

percents

A matrix or a list of matrices (depending on the number of factors of the dependent variable) with regression coefficients expressed as percentages

boot.ci

A matrix or a list of matrices (depending on the number of factors of the dependent variable) with confidence intervals computed with bootstrap

Details

This function works with binomial as well as multinomial regression models. If the dependant variable has two factors, glm is used, if it has more than two factors multinom function (from nnet package) is used. The function expresses the regression coefficients as percentages through three distinct methods: raw percentages, 'pure effects' percentages and 'experimental effects' percentages (see Deauvieau, 2010).

References

Deauvieau, J. (2010), 'Comment traduire sous forme de probabilites les resultats d'une modelisation logit ?', Bulletin of Sociological Methodology / Bulletin de Methodologie Sociologique 105(1), 5-23.

Deauvieau, J. (2011), 'Est-il possible et souhaitable traduire sous forme de probabilites un coefficient logit ? Reponse aux remarques formulees par Marion Selz a propos de mon article paru dans le BMS en 2010', Bulletin of Sociological Methodology / Bulletin de Methodologie Sociologique 112(1), 32-42.

See Also

glm, multinom

Examples

# NOT RUN {
  ## An example for binomial logit regression
  data(Music)
  translate.logit(Daily ~ Gender + Age, Music)
  
  ## An example for multinomial logit regression
  translate.logit(OnlyMus ~ Gender + Age, Music)
# }