[R] Fwd: Categorial Response Questions

[Simon Knapp]

Not the same data your using but...

# get something like a reasonable
# age dist for dummy data:
data('mw.ages', package='UsingR')
mw.age <- mw.ages[2:80,]

# size of dummy dataset
n.obs <- 1000

#pr of survival
pr.s <- 0.80

# dummy data
dat <- data.frame(
    pclass=sample(c('1st', '2nd', '3rd'), n.obs, T),
    age=sample(1:79, n.obs, T, apply(mw.age, 1, sum)/sum(mw.age)),
    sex=sample(c('male', 'female'), n.obs, T, apply(mw.age, 2,
sum)/sum(mw.age)),
    survived=sample(0:1, n.obs, T, c(1-pr.s, pr.s))
)
dat$age.group <- cut(dat$age, seq(0, 80, 10), right=F)


#-----------------------------------------------------
# get the logits
attach(dat)
logits <- by(dat, list(age.group, sex, pclass), function(x){
    p <- sum(x$survived)/nrow(x)
    p/(1-p)
})
detach(dat)
#-----------------------------------------------------
#-----------------------------------------------------
# the class of the object logits is 'by' -
# this function munges it into a data frame
# (note that 'as.data.frame' is generic)

# ARGS
# x               an object of class 'by'
# out.names names you want to use
#                 in the output data frame.

as.data.frame.by <- function(x, out.names=NULL) {
    nms <- dimnames(x)
    no.levs <- sapply(nms, length)
    res <- as.numeric(x)
    for(i in 1:length(nms)){
        res <- data.frame(
            if(i==1)
                rep(nms[[i]], times=prod(no.levs[(i+1):length(nms)]))
            else if(i==length(nms))
                rep(nms[[i]], each=prod(no.levs[1:(i-1)]))
            else
                rep(nms[[i]], each=prod(no.levs[1:(i-1)]),
times=prod(no.levs[(i+1):length(nms)]))
        , res)
    }
    names(res) <- out.names
    res
}

as.data.frame(logits, c('age.group', 'sex', 'pclass', 'logit'))
#-----------------------------------------------------





2008/10/18 andyer weng <nzandyer at gmail.com>:
> hi all,
>
> For my question in the first email below, I found I made a mistake on
> my coding in the previous email, the one I was trying to type should
> be
>
>> grouped.titanic.df<-data.frame(group.age.group=sort(unique(titanic.df$age.group)),
> + expand.grid(sex=sort(unique(titanic.df$sex)),pclass=sort(unique(titanic.df$pclass))),
> + r=as.vector(tapply(titanic.df$survived,titanic.df$age.group,sum)),
> + n=as.vector(tapply(titanic.df$survived,titanic.df$age.group,length)))
>
> Error in data.frame(group.age.group = sort(unique(titanic.df$age.group)),  :
>  arguments imply differing number of rows: 8, 6
>
>
> please advise what I have done wrong? why the error message come up.
> Am I doing the right thing to fix the question i mentioned in the
> first email (the bottom email)?
>
> Cheers. Andyer
>
>
> ---------- Forwarded message ----------
> From: andyer weng <nzandyer at gmail.com>
> Date: 2008/10/18
> Subject: Fwd: Categorial Response Questions
> To: r-help at r-project.org
>
>
> hi all,
>
> me again. i try to type the following coding for my question below,
> but it comes up a error messgae. please advise whether the way i was
> trying to do will solve my question stated in the previous email. If
> so , please advise what is wrong with my coding.
> (p.s. all the data are stored in xxx.df)
>
>>  grouped.xxx.df<-data.frame(group.age.group=sort(unique(xxx.df$age.group)),
> + expand.grid(sex=c("female","male"),age.group=c("0-9","10-19","20-29","30-39","40-49","50-59","60-69","70-79")),
> + r=tapply(xxx.df$survived,titanic.df$age.group,sum),
> + n=tapply(xxx.df$survived,titanic.df$age.group,length))
>
> Error in data.frame(group.age.group = sort(unique(xxx.df$age.group)),  :
>  arguments imply differing number of rows: 8, 16
> In addition: Warning messages:
> 1: In Ops.factor(left) : + not meaningful for factors
> 2: In Ops.factor(left) : + not meaningful for factors
>
>
> thanks millions.
>
> Regards,
> Andyer
>
>
>
>
>
>
> ---------- Forwarded message ----------
> From: andyer weng <nzandyer at gmail.com>
> Date: 2008/10/18
> Subject: Fwd: Categorial Response Questions
> To: r-help at r-project.org
>
>
> Sorry Guys, i press the wrong button to send out the uncompleted message.
>
> let me do it again.
>
> my purpose for below questions  is to assess the effect of class, age
> and sex on the survival.
>
>
> I have a data set containing :
>
> pclass:  A factor giving the class of the passenger: one of 1st, 2nd, 3rd.
> age:      The age of the passenger in years.
> sex:      Passenger's gender: female or male
> age.group:        Passengers age group, one of 0‐9 , 10‐19, 20‐29,
> 30‐39, 40‐49, 50‐59, 60‐69,70‐79
> survived:            Passenger's survival (1=survived, 0=did not survive)
>
> Ignoring the variable age,
> - I need to group the data into groups corresponding to each
> age‐group/sex/class combination,
> - I need to compute the logits for each combination.
> - Make a data frame containing the logits, and the categorical
> variables. I need to have one line in the data frame for each
> combination of the factor levels.
>
> Can someone please help with the R code for above???!!!
>
> Thanks millions!!
>
> Cheers
> Andyer.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>