如何减少 x 轴上两个离散值之间的空间？

Question

首先，我想在没有 ggplot 的情况下做到这一点。

我有一个用于箱线图的脚本：

data <- data.frame(
  u    = c(0.522, 0.488, 0.474, 0.443, 0.510, 0.443, 0.420, 0.554, 0.333, 0.414,
           0.467, 0.606, 0.588),
  ub   = c(0.594, 0.568, 0.578, 0.523, 0.577, 0.501, 0.522, 0.623, 0.498, 0.511,
           0.544, 0.654, 0.639),
  nu   = c(0.646, 0.620, 0.644, 0.607, 0.667, 0.614, 0.631, 0.701, 0.586, 0.589,
           0.664, 0.739, 0.744),
  nub  = c(0.733, 0.701, 0.700, 0.698, 0.724, 0.701, 0.722, 0.777, 0.630, 0.663,
           0.703, 0.78, 0.787),
  wnub = c(0.815, 0.782, 0.757, 0.764, 0.792, 0.771, 0.838, 0.860, 0.700, 0.703,
           0.774, 0.837, 0.863))
boxplot(data, las=1, par(mar=c(7.5, 4.2, 0.25, 0.15)), axes=FALSE, xlab=NA,
        ylab=NA, boxwex=.4, outline=FALSE, ylim=c(0:1))
box()
labnames = c("Unigrams",
             "Unigrams\n& Bigrams",
             "Noun Unigrams",
             "Noun Unigrams\n& Noun Bigrams",
             "Noun Unigrams\n& Noun Bigrams\n(positional weights)")
axis(side=1, tck=-0.025, labels=FALSE)
axis(side=1, lwd=0, labels=FALSE, at=c(1:5), cex.axis=1)
text(seq(1, 5, by=1), par("usr")[3]-.2, labels=labnames, srt=-90,
     xpd=TRUE, adj=.25)
axis(side=2, tck=-0.02, labels=NA)
axis(side=2, lwd=0, cex.axis=1, las=1)
mtext(side=1, "Feature Sets", line=6.5, cex=1.5)
mtext(side=2, "F-Score", line=3, cex=1.5)

产生下图：

我对浪费的空白数量感到恼火。我一直试图减少空间的一种方法是使用“asp”将离散的 x 轴值靠得更近，但它不会改变任何东西。我怎样才能做到这一点，以便减少盒子之间的巨大差距？

另外，如果有任何其他方法可以减少此图片中的空白，请分享您的想法。

Answer 1

Option 1: If you're saving images as PDF for Latex, you can set the width and height with pdf(file=file, width=width, height=height). Then you can ignore the boxwex parameter altogether.

Option 2: Alternatively, you can use the at parameter to tell boxplot where in the X axis it should place the boxes:

at.x <- seq(1,by=.5,length.out=5) # set here the X-axis positions
boxplot(data, par(mar=c(8.5, 4.2, 0.25, 0.15)), outline=FALSE,
        cex.lab=1.5, ylab="F-score", xlab=NA, las=3, names=labnames,
        boxwex=.4, ylim=c(0:1),
        at=at.x)
# display the X-axis name
mtext(side=1, "Feature Sets", line=7.2, cex=1.5)

See that the code is also heavily simplified. You can further play with the xlim parameter to boxplot to reduce the inner margins of the plot. Try for instance xlim=c(0.8,3.2). Result:

选项 3：如果您确实想节省空间，则可以使用颜色和图例：

cols <- c("white","red","yellow","green","blue")
boxplot(data, par(mar=c(0.2, 3.2, 0.25, 0.15)), outline=FALSE,
        axes=F, ylab=NA, xlab=NA, names=NA, col=cols)
box()
axis(2)
mtext(side=2,"F-score",line=2.2)
legend("bottomright", cex=.8, fill=cols,
       c("Unigrams", "Unigrams & Bigrams", "Noun Unigrams",
         "Noun Unigrams & Noun Bigrams",
         "Noun Unigrams & Noun Bigrams (positional weights)"))

最后，为什么将 Y 轴的限制设置为 0 和 1？那在那里浪费了很多空间。另外，为什么要删除异常值？只是作为一个 IR 研究员问自己 :-)