% \VignetteIndexEntry{Birth Data - Bivariate Binary Regression}
% \VignetteDepends{VGAM}
%\VignetteEngine{knitr::knitr}
%\VignetteEncoding{UTF-8}
\documentclass[a4paper]{article}

\title{Birth Data - Bivariate Binary Regression}

\begin{document}

\maketitle


<<echo=FALSE,eval=FALSE>>=
options(width=80)
@


First the Birth data are loaded from package "catdata".

<<results='hide',eval=FALSE>>=
library(catdata)
data(birth)
attach(birth)
@

Now  the original variable "Intensive" is converted into the binary variable "Intensive" indicating 
whether the child spent time in intensive care or not.

<<eval=FALSE>>=
intensive <- rep(0,length(Intensive))
intensive[Intensive>0] <- 1
Intensive <- intensive
@

Now "Previous" is reduced to 3 categories by merging two and more previous pregnancies to level "2".

<<eval=FALSE>>=
previous <- Previous
previous[previous>1] <- 2 
Previous <- previous
@

<<eval=FALSE>>=
library(VGAM)
@

The data set "Birth" is built as data set containing the variables for the model but without  missing values.

<<eval=FALSE>>=
Birth <- as.data.frame(na.omit(cbind(Intensive, Cesarean, Sex, Weight, Previous, 
AgeMother)))
detach(birth)
@

With that data set the model can be fitted. The option "binom2.or" is needed to fit a bivariate binary model. 

<<eval=FALSE>>=
bivarlogit <- vglm(cbind(Intensive , Cesarean) ~ as.factor(Sex) + Weight + 
as.factor(Previous) + AgeMother, binom2.or(zero=NULL), data=Birth)

summary(bivarlogit)
@

<<echo=FALSE,eval=FALSE>>=
detach(package:VGAM)
@

\end{document}