Question

Edit: Building off of aL3xa’s answer below, I’ve modified his syntax below. Not perfect, but getting closer. I still haven’t found a way to make xtable accept \multicolumn{} arguments for columns or rows. It also appears that Hmisc handles some of these type of tasks behind the scenes, but it looks like a bit of an undertaking to understand what’s going on there. Does anyone have experience with the latex function in Hmisc?

ctab <- function(tab, dec = 2, margin = NULL) {
    tab <- as.table(tab)
    ptab <- paste(round(prop.table(tab, margin = margin) * 100, dec), "%", sep = "")
    res <- matrix(NA, nrow = nrow(tab) , ncol = ncol(tab) * 2, byrow = TRUE)
    oddc <- 1:ncol(tab) %% 2 == 1
    evenc <- 1:ncol(tab) %% 2 == 0
    res[,oddc ] <- tab
    res[,evenc ] <- ptab
    res <- as.table(res)
    colnames(res) <- rep(colnames(tab), each = 2)
    rownames(res) <- rownames(tab)
    return(res)
}

I would like to create a table formatted for LaTeX output that contains both the counts and percentages for each column or variable. I have not found a ready made solution to this problem, but feel I must be recreating the wheel to some extent.

I have developed a solution for straight tabulations, but am struggling with adopting something for a cross tabulation.

First some sample data:

#Generate sample data
dow <- sample(1:7, 100, replace=TRUE)
purp <- sample(1:4, 100, replace=TRUE)
dow <- factor(dow, 1:7, c("Mon", "Tues", "Wed", "Thurs", "Fri", "Sat", "Sun"))
purp <- factor(purp, 1:4, c("Business", "Commute", "Vacation", "Other"))

And now the working straight tab function:

customTable <- function(var, capt = NULL){
    counts <- table(var)
    percs <- 100 * prop.table(counts)       

    print(
        xtable(
            cbind(
                Count = counts
                , Percent = percs
            )
        , caption = capt
        , digits = c(0,0,2)
        )
    , caption.placement="top"
    )
}

#Usage
customTable(dow, capt="Day of Week")
customTable(purp, capt="Trip Pupose")

Does anyone have any suggestions for adopting this for cross tabulations (i.e. day of week BY trip purpose)? Here is what I’ve currently written, which does NOT use the xtable library and ALMOST works, but is not dynamic and is quite ugly to work with:

#Create table and percentages
a <- table(dow, purp)
b <- round(prop.table(a, 1),2)

#Column bind all of the counts & percentages together, this SHOULD become dynamic in future
d <- cbind( cbind(Count = a[,1],Percent =  b[,1])
        , cbind(Count = a[,2], Percent = b[,2])
        , cbind(Count = a[,3], Percent = b[,3])
        , cbind(Count = a[,4], Percent = b[,4])
)

#Ugly function that needs help, or scrapped for something else
crossTab <- function(title){
    cat("\\begin{table}[ht]\n")
    cat("\\begin{center}\n")
    cat("\\caption{", title, "}\n", sep="") 

    cat("\\begin{tabular}{rllllllll}\n")
    cat("\\hline\n")

    cat("", cat("", paste("&\\multicolumn{2}{c}{",colnames(a), "}"), sep = ""), "\\\\\n", sep="")
    c("&", cat("", colnames(d), "\\\\\n", sep=" & "))
    cat("\\hline\n")
    c("&", write.table(d, sep = " & ", eol="\\\\\n", quote=FALSE, col.names=FALSE))

    cat("\\hline\n")
    cat("\\end{tabular}\n")
    cat("\\end{center}\n")
    cat("\\end{table}\n")   
}   

crossTab(title = "Day of week BY Trip Purpose")

Answer 1

I wasn’t able to figure out how to generate a multi column header using xtable, but I did realize that i could concatenate my counts & percentages into the same column for printing purposes. Not ideal, but seems to get the job done. Here’s the function I’ve written:

ctab3 <- function(row, col, margin = 1, dec = 2, percs = FALSE, total = FALSE, tex = FALSE, caption = NULL){
    tab <- as.table(table(row,col))
    ptab <- signif(prop.table(tab, margin = margin), dec)

    if (percs){

        z <- matrix(NA, nrow = nrow(tab), ncol = ncol(tab), byrow = TRUE) 
        for (i in 1:ncol(tab)) z[,i] <- paste(tab[,i], ptab[,i], sep = " ")
        rownames(z) <- rownames(tab)
        colnames(z) <- colnames(tab)

        if (margin == 1 & total){
            rowTot <- paste(apply(tab, 1, sum), apply(ptab, 1, sum), sep = " ")
            z <- cbind(z, Total = rowTot)
        } else if (margin == 2 & total) {
            colTot <- paste(apply(tab, 2, sum), apply(ptab, 2, sum), sep = " ")
            z <- rbind(z,Total = colTot)
        }
    } else {
        z <- table(row, col)    
    }
ifelse(tex, return(xtable(z, caption)), return(z))
}

Probably not the final product, but does allow for some flexibility in parameters. At the most basic level, is only a wrapper of table() but can also generate LaTeX formatted output as well. Here is what I ended up using in a Sweave document:

<<echo = FALSE>>=
for (i in 1:ncol(df)){
    print(ctab3(
        col = df[,1]
        , row = df[,i]
        , margin = 2
        , total = TRUE
        , tex = TRUE
        , caption = paste("Dow by", colnames(df[i]), sep = " ")
    ))
}
@