Error in plot.BsProb in R when trying to change xlab

Question

I am using the package BsMD to plot marginal factor posterior probabilities for Bayesian screening (plot.BsProb). I wish to change the xlab and ylab, but I get an error:

library(BsMD) 

X = matrix(c(1, 1, 1, 1, 1,-1,-1,-1,-1,-1,-1, 1, 1, 1,-1,-1,-1,-1,-1,
             1,-1, 1,-1,-1, 1,-1, 1,-1, 1,-1,-1,-1,-1,-1, 1,-1,-1,-1,
            -1,-1, 1, 1, 1, 1, 1,-1,-1,-1, 1,-1, 1, 1, 1, 1,-1,-1, 1,
             1, 1, 1,-1,-1, 1, 1, 1,-1,-1,-1, 1,-1,-1, 1, 1, 1,-1,-1,
             1, 1,-1, 1,-1, 1,-1,-1, 1,-1, 1, 1, 1,-1, 1,-1, 1, 1,-1
             ),nrow=19,ncol=5)    
y = matrix(c(6,2,4,11,17,4,2,6,18,11,10,12,12,15,5,9,11,10,9),nrow=19,ncol=1)

Models = BsProb(X = X, y = y, blk = 0, mFac = 5, mInt = 1, p = 0.25, g = 2.5, ng = 1, nMod = 5)
plot(Models, code=FALSE, xlab="asdasdasdasd")

Error in plot.default(x, y[, 1], xlim = range(x), ylim = c(0, 1), type = "n", : formal argument "xlab" matched by multiple actual arguments.

May anyone please help me change the xlab and ylab for plotting "Models"?

Answer 1

You could temporarily change the par settings for label colors and add the labels afterwards using title .

## create plot with white axis labels
par(col.lab = "white")
plot(Models, code = FALSE)

## add customized axis labels
par(col.lab = "black")
title(xlab = "X-axis label", ylab = "Y-axis label") 

Answer 2

Good old traceback() tells us what's happening here:

plot(Models, code=FALSE, xlab="asdasdasdasd");
## Error in plot.default(x, y[, 1], xlim = range(x), ylim = c(0, 1), type = "n",  :
##   formal argument "xlab" matched by multiple actual arguments
traceback();
## 5: plot.default(x, y[, 1], xlim = range(x), ylim = c(0, 1), type = "n",
##        xlab = "factors", ylab = "posterior probability", frame = FALSE,
##        axes = FALSE, ...)
## 4: plot(x, y[, 1], xlim = range(x), ylim = c(0, 1), type = "n",
##        xlab = "factors", ylab = "posterior probability", frame = FALSE,
##        axes = FALSE, ...)
## 3: spikes(prob, ...)
## 2: plot.BsProb(Models, code = FALSE, xlab = "asdasdasdasd")
## 1: plot(Models, code = FALSE, xlab = "asdasdasdasd")

Your Models object is S3-classed as BsProb :

class(Models);
## [1] "BsProb" "list"

Thus running plot(Models,...) dispatches to plot.BsProb() which exists under the BsMD private environment:

BsMD:::plot.BsProb;
## function (x, code = TRUE, prt = FALSE, cex.axis = par("cex.axis"),
##     ...)
## {
##     spikes <- function(prob, lwd = 3, ...) {
##         y <- prob
##         n <- nrow(y)
##         x <- seq(n)
##         lab <- rownames(prob)
##         plot(x, y[, 1], xlim = range(x), ylim = c(0, 1), type = "n",
##             xlab = "factors", ylab = "posterior probability",
##             frame = FALSE, axes = FALSE, ...)
##         if (ncol(y) == 1) {
##             for (i in x) segments(x[i], 0, x[i], y[i, 1], lwd = lwd,
##                 col = grey(0.2))
##         }
##         else {
##             y[, 1] <- apply(prob, 1, min)
##             y[, 2] <- apply(prob, 1, max)
##             for (i in x) {
##                 segments(x[i], 0, x[i], y[i, 2], lwd = lwd, col = grey(0.8),
##                   lty = 1)
##                 segments(x[i], 0, x[i], y[i, 1], lwd = lwd, col = grey(0.2),
##                   lty = 1)
##             }
##         }
##         axis(1, at = x, labels = lab, line = 0, cex.axis = cex.axis)
##         axis(2, cex.axis = cex.axis)
##         invisible(NULL)
##     }
##     if (!any(class(x) == "BsProb"))
##         return("\nArgument `x' should be class BsProb. Output of corresponding function.")
##     ifelse(x$INDGAM == 0, prob <- as.matrix(x$sprob), prob <- x$prob)
##     if (code)
##         rownames(prob) <- rownames(x$prob)
##     else rownames(prob) <- names(x$sprob)
##     spikes(prob, ...)
##     if (prt)
##         summary.BsProb(x)
##     invisible(NULL)
## }
## <environment: namespace:BsMD>

Looking into the above code, we can see that it dynamically defines a spikes() function and calls it near the end of the body. The spikes() function calls plot() and unconditionally passes an argument of xlab='factors' , along with the ... variadic arguments relayed from the top-level call. That's why there are two xlab actual arguments matching against the xlab formal argument; one comes from the spikes() lexical argument list, and one comes from the ... relay.

Thus, unfortunately, there's no way to pass xlab from the top-level call without clashing with the implementation. You're out of luck. You can write a strongly-worded letter to the maintainer of the BsMD package requesting they make the BsMD:::plot.BsProb() function more flexible, and in the meantime just bite the bullet and use a hack to work around the limitation, like @fdetsch's excellent suggestion.