crosstab for n-way contingency tables

Dear list.
 
New to R, I'm looking for a way of using crosstab to output
low-dimensional (higher than 2) contingency tables (frequencies,
per-cents by rows, % by columns, mean, quantiles....) I'm looking for
something of the following sort
 
dataframe: singers, 
categorical variates: voice category (soprano,mezzo-soprano, ...) ,
voice type( drammatic, spinto, lirico-spinto, lirico, leggero), school
(german, italian, french, russian, anglo-saxon, other);repertory
(opera, Lieder, oratorio, operetta)
continuous variate: age
 
I would like to tabulate the frequencies (relative percentages)  say
in the following way
columns: school  repertory
rows :      voice category  voice type
 
or to output in the cells of the above table, the statistics
(mean/median/quantiles) for age
 

I've seen that the function bwplot(age~school | repertory, data=
singers, layout=c(4,2)) would do graphically something similar to what
I want, but I desire the output also in tabular form
 
Thanks
 
Isotta

I don't know that you will find a single function that will do all of
what you desire, but you can look at the ctab() function, which is in
the 'catspec' package on CRAN by John Hendrickx. This will do multi-way
tables with summary row/col statistics.

There is also the CrossTable() function in the 'gmodels' package on
CRAN, though this will only do one way and two way tables, with summary
row/col statistics.

Neither of the above provide typical summary output for continuous data.
They are more for categorical variables.

For simple count multi-way output, you can also look at the ftable()
function which is in the base package. Also look at the summary()
function in the base package which will provide range, mean, quantile
data for continuous variables. It may just be a matter of formatting the
output in a fashion that you desire on a post analysis basis.

HTH,

Marc Schwartz

Hi Isotta,

You can do this with the reshape package (available from CRAN). eg

install.packages("reshape")
library(reshape)

data(singer, package="lattice")
singer$type <- c("drammatic", "spinto", "lirico-spinto", "lirico",
"leggero")[sample(1:5, 235, replace=T)]
singer$school <- c("german", "italian", "french", "russian",
"anglo-saxon", "other")[sample(1:6, 235, replace=T)]
singer$repertory <- c("opera", "Lieder", "oratorio",
"operetta")[sample(1:4, 235, replace=T)]

# First deshape the data (this puts it in a form easy to reshape)
singer_d <- deshape(singer, id=c("type", "school", "repertory",
"voice.part"), m="height")

# You can do things like
reshape(singer_d, type  ~ school, length, subset=variable=="height")
reshape(singer_d, type  ~ school, mean, subset=variable=="height")

# or with margins
reshape(singer_d, type + school ~ ., length,
subset=variable=="height", margins=c("type","grand_row"))

# or with multiple stats
opera.sum <- function(x) c(min=min(x), mean=mean(x), max=max(x))
reshape(singer_d, type + school ~ ., opera.sum, subset=variable=="height")

# What you'd want for your data, but doesn't work well with this example
# and is going to be a big table regardless!
reshape(singer_d, type + voice.part ~ school + repertory, length)

There's some more info available at http://had.co.nz/reshape but I'm
still working on it.

Hadley