使用 R.zoo 绘制带有误差线的多个系列

Question

我的数据如下所示：

   > head(data)
             groupname ob_time dist.mean  dist.sd dur.mean   dur.sd   ct.mean    ct.sd
      1      rowA     0.3  61.67500 39.76515 43.67500 26.35027  8.666667 11.29226
      2      rowA    60.0  45.49167 38.30301 37.58333 27.98207  8.750000 12.46176
      3      rowA   120.0  50.22500 35.89708 40.40000 24.93399  8.000000 10.23363
      4      rowA   180.0  54.05000 41.43919 37.98333 28.03562  8.750000 11.97061
      5      rowA   240.0  51.97500 41.75498 35.60000 25.68243 28.583333 46.14692
      6      rowA   300.0  45.50833 43.10160 32.20833 27.37990 12.833333 14.21800

每个组名都是一个数据系列。由于我想分别绘制每个系列，因此我将它们分开：

> A <- zoo(data[which(groupname=='rowA'),3:8],data[which(groupname=='rowA'),2])
> B <- zoo(data[which(groupname=='rowB'),3:8],data[which(groupname=='rowB'),2])
> C <- zoo(data[which(groupname=='rowC'),3:8],data[which(groupname=='rowC'),2])

ETA：

Thanks to gd047: Now I'm using this:

    z <- dlply(data,.(groupname),function(x) zoo(x[,3:8],x[,2]))

生成的动物园对象如下所示：

> head(z$rowA)
          dist.mean  dist.sd dur.mean   dur.sd   ct.mean    ct.sd
     0.3  61.67500 39.76515 43.67500 26.35027  8.666667 11.29226
     60   45.49167 38.30301 37.58333 27.98207  8.750000 12.46176
     120  50.22500 35.89708 40.40000 24.93399  8.000000 10.23363
     180  54.05000 41.43919 37.98333 28.03562  8.750000 11.97061
     240  51.97500 41.75498 35.60000 25.68243 28.583333 46.14692
     300  45.50833 43.10160 32.20833 27.37990 12.833333 14.21800

因此，如果我想根据时间绘制 dist.mean 并包含等于 +/- dist.sd 的误差线对于每个系列：

• 如何组合 A、B、C dist.mean 和 dist.sd？
• 如何制作条形图，或者更好，结果对象的折线图？

Answer 1

我不认为将数据分成三部分只是为了将它们组合在一起以绘制图表有什么意义。这是使用ggplot2库绘制的图：

library(ggplot2)
qplot(ob_time, dist.mean, data=data, colour=groupname, geom=c("line","point")) + 
  geom_errorbar(aes(ymin=dist.mean-dist.sd, ymax=dist.mean+dist.sd))

它沿着自然比例间隔时间值，您可以使用scale_x_continuous来定义实际时间值处的刻度线。让它们等间距比较棘手：您可以将 ob_time 转换为一个因子，但 qplot 拒绝用线连接这些点。

解决方案 1 - 条形图：

qplot(factor(ob_time), dist.mean, data=data, geom=c("bar"), fill=groupname, 
      colour=groupname, position="dodge") + 
geom_errorbar(aes(ymin=dist.mean-dist.sd, ymax=dist.mean+dist.sd), position="dodge")

解决方案 2 - 使用因子的 1,2,... 重新编码手动添加线条：

qplot(factor(ob_time), dist.mean, data=data, geom=c("line","point"), colour=groupname) +
  geom_errorbar(aes(ymin=dist.mean-dist.sd, ymax=dist.mean+dist.sd)) + 
  geom_line(aes(x=as.numeric(factor(ob_time))))

Answer 2

这是我尝试这样做的方式的提示。我忽略了分组，因此您必须修改它以包含多个系列。我也没有去过动物园，因为我了解不多。

g <- (nrow(data)-1)/(3*nrow(data))

plot(data[,"dist.mean"],col=2, type='o',lwd=2,cex=1.5, main="This is the title of the graph",
 xlab="x-Label", ylab="y-Label", xaxt="n",
 ylim=c(0,max(data[,"dist.mean"])+max(data[,"dist.sd"])),
 xlim=c(1-g,nrow(data)+g))
axis(side=1,at=c(1:nrow(data)),labels=data[,"ob_time"])

for (i in 1:nrow(data)) {
lines(c(i,i),c(data[i,"dist.mean"]+data[i,"dist.sd"],data[i,"dist.mean"]-data[i,"dist.sd"]))
lines(c(i-g,i+g),c(data[i,"dist.mean"]+data[i,"dist.sd"], data[i,"dist.mean"]+data[i,"dist.sd"]))
lines(c(i-g,i+g),c(data[i,"dist.mean"]-data[i,"dist.sd"], data[i,"dist.mean"]-data[i,"dist.sd"]))
}

Answer 3

使用带有 split= 参数的 read.zoo 读取数据，以按组名拆分数据。然后将距离、下线和上线绑在一起。最后绘制它们。

Lines <- "groupname ob_time dist.mean  dist.sd dur.mean   dur.sd   ct.mean    ct.sd
rowA     0.3  61.67500 39.76515 43.67500 26.35027  8.666667 11.29226
rowA    60.0  45.49167 38.30301 37.58333 27.98207  8.750000 12.46176
rowA   120.0  50.22500 35.89708 40.40000 24.93399  8.000000 10.23363
rowA   180.0  54.05000 41.43919 37.98333 28.03562  8.750000 11.97061
rowB   240.0  51.97500 41.75498 35.60000 25.68243 28.583333 46.14692
rowB   300.0  45.50833 43.10160 32.20833 27.37990 12.833333 14.21800"

library(zoo)
# next line is only needed until next version of zoo is released
source("http://r-forge.r-project.org/scm/viewvc.php/*checkout*/pkg/zoo/R/read.zoo.R?revision=719&root=zoo")
z <- read.zoo(textConnection(Lines), header = TRUE, split = 1, index = 2)

# pick out the dist and sd columns binding dist with lower & upper 
z.dist <- z[, grep("dist.mean", colnames(z))]
z.sd <- z[, grep("dist.sd", colnames(z))]
zz <- cbind(z = z.dist, lower = z.dist - z.sd, upper = z.dist + z.sd)

# plot using N panels
N <- ncol(z.dist)
ylab <- sub("dist.mean.", "", colnames(z.dist))
plot(zz, screen = 1:N, type = "l", lty = rep(1:2, N*1:2), ylab = ylab)

Answer 4

我认为您不需要为这种类型的图创建动物园对象，我会直接从数据框中创建动物园对象。当然，可能还有其他原因需要使用 Zoo 对象，例如智能合并、聚合等。

一种选择是latticeExtra 中的 segplot 函数，

library(latticeExtra)
segplot(ob_time ~ (dist.mean + dist.sd) + (dist.mean - dist.sd) | groupname, 
    data = data, centers = dist.mean, horizontal = FALSE)
## and with the latest version of latticeExtra (from R-forge):
trellis.last.object(segments.fun = panel.arrows, ends = "both", angle = 90, length = .1) +
    xyplot(dist.mean ~ ob_time | groupname, data, col = "black", type = "l")

使用 Gabor 的可重现性良好的数据集会生成：