ggm (version 2.2)

fitmlogit: Multivariate logistic models

Description

Fits a logistic regression model to multivariate binary responses.

Usage

fitmlogit(..., C = c(), D = c(), data, mit = 100, ep = 1e-80, acc = 1e-04)

Arguments

...

Model formulae of marginal logistic models for each response and for each association parameters (log-odds ratios).

C

Matrix of equality constraints.

D

Matrix of inequality cosntraints.

data

A data frame containing the responses and the explanatory variables.

mit

A positive integer: maximum number of iterations. Default: 100.

ep

A tolerance used in the algorithm: default 1e-80.

acc

A tolerance used in the algorithm: default 1e-4.

Value

• LL
  The maximized log-likelihood.
• be
  The vector of the Maximum likelihood estimates of the parameters.
• S
  The estimated asymptotic covariance matrix.
• P
  The estimated cell probabilities for each individual.

Details

See Evans and Forcina (2011).

References

Evans, R.J. and Forcina, A. (2011). Two algorithms for fitting constrained marginal models. Submitted, arXiv:1110.2894v1 [stat.CO].

See Also

glm

Examples

data(surdata)                     
out1 <- fitmlogit(A ~X, B ~ Z, cbind(A, B) ~ X*Z, data = surdata)     
out1$beta
out2 <- fitmlogit(A ~X, B ~ Z, cbind(A, B) ~ 1, data = surdata)        
out2$beta