创建格子化（刻面）薄板样条响应曲面

Question

我试图绘制一组薄板样条响应曲面，用于与两个连续变量加一个离散变量相关的测量。到目前为止，我已经基于离散变量对数据进行了子集化以生成图对，但在我看来应该有一种方法来创建一些光滑的格子图。似乎可以通过ggplot2与geom_tile和geom_contour在ggplot2中制作热图来完成此操作，但我仍然坚持

（1）如何重新组织数据（或解释预测的表面数据）以便用rsm绘图？

（2）使用基本图形创建格子化热图的语法？或

（3）使用rsm中的图形来实现此目的的方法（library(fields) library(ggplot2) sumframe<-structure(list(Morph = c("LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "LW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW", "SW"), xvalue = c(4, 8, 9, 9.75, 13, 14, 16.25, 17.25, 18, 23, 27, 28, 28.75, 4, 8, 9, 9.75, 13, 14, 16.25, 17.25, 18, 23, 27, 28, 28.75), yvalue = c(17, 34, 12, 21.75, 29, 7, 36.25, 14.25, 24, 19, 36, 14, 23.75, 17, 34, 12, 21.75, 29, 7, 36.25, 14.25, 24, 19, 36, 14, 23.75), zvalue = c(126.852666666667, 182.843333333333, 147.883333333333, 214.686666666667, 234.511333333333, 198.345333333333, 280.9275, 246.425, 245.165, 247.611764705882, 266.068, 276.744, 283.325, 167.889, 229.044, 218.447777777778, 207.393, 278.278, 203.167, 250.495, 329.54, 282.463, 299.825, 286.942, 372.103, 307.068)), .Names = c("Morph", "xvalue", "yvalue", "zvalue"), row.names = c(NA, -26L), class = "data.frame") sumframeLW<-subset(sumframe, Morph=="LW") sumframeSW<-subset(sumframe, Morph="SW") split.screen(c(1,2)) screen(n=1) surf.teLW<-Tps(cbind(sumframeLW$xvalue, sumframeLW$yvalue), sumframeLW$zvalue, lambda=0.01) summary(surf.teLW) surf.te.outLW<-predict.surface(surf.teLW) image(surf.te.outLW, col=tim.colors(128), xlim=c(0,38), ylim=c(0,38), zlim=c(100,400), lwd=5, las=1, font.lab=2, cex.lab=1.3, mgp=c(2.7,0.5,0), font.axis=1, lab=c(5,5,6), xlab=expression("X value"), ylab=expression("Y value"),main="LW plot") contour(surf.te.outLW, lwd=2, labcex=1, add=T) points(sumframeLW$xvalue, sumframeLW$yvalue, pch=21) abline(a=0, b=1, lty=1, lwd=1.5) abline(a=0, b=1.35, lty=2) screen(n=2) surf.teSW<-Tps(cbind(sumframeSW$xvalue, sumframeSW$yvalue), sumframeSW$zvalue, lambda=0.01) summary(surf.teSW) surf.te.outSW<-predict.surface(surf.teSW) image(surf.te.outSW, col=tim.colors(128), xlim=c(0,38), ylim=c(0,38), zlim=c(100,400), lwd=5, las=1, font.lab=2, cex.lab=1.3, mgp=c(2.7,0.5,0), font.axis=1, lab=c(5,5,6), xlab=expression("X value"), ylab=expression("Y value"),main="SW plot") contour(surf.te.outSW, lwd=2, labcex=1, add=T) points(sumframeSW$xvalue, sumframeSW$yvalue, pch=21) abline(a=0, b=1, lty=1, lwd=1.5) abline(a=0, b=1.35, lty=2) close.screen(all.screens=TRUE) 可以应对高阶曲面，所以我可以在某种程度上强制执行，但是情节没有完全格式化） 。

这是我到目前为止一直在使用的一个例子：

{{1}}

Answer 1

如评论中所述，melt()可用于重塑Tps()输出，然后可以重新格式化（删除NA），重新组合成单​​个数据帧并绘制。以下是ggplot2和levelplot：

的图表

library(reshape)
library(lattice)

LWsurfm<-melt(surf.te.outLW)
LWsurfm<-rename(LWsurfm, c("value"="z", "Var1"="x", "Var2"="y"))
LWsurfms<-na.omit(LWsurfm)
SWsurfms[,"Morph"]<-c("SW")

SWsurfm<-melt(surf.te.outSW)
SWsurfm<-rename(SWsurfm, c("value"="z", "X1"="x", "X2"="y"))
SWsurfms<-na.omit(SWsurfm)
LWsurfms[,"Morph"]<-c("LW")

LWSWsurf<-rbind(LWsurfms, SWsurfms)

LWSWp<-ggplot(LWSWsurf, aes(x,y,z=z))+facet_wrap(~Morph)
LWSWp<-LWSWp+geom_tile(aes(fill=z))+stat_contour()
LWSWp

或：     levelplot（z~x * y | Morph，data = LWSWsurf，contour = TRUE）

Answer 2

require(rgl)
open3d()
plot3d
surface3d(surf.te.outSW$x, surf.te.outSW$y, surf.te.outSW$z, col="red")
surface3d(surf.te.outLW$x, surf.te.outLW$y, surf.te.outLW$z, col="blue")
decorate3d()
      rgl.snapshot("OutRGL.png")

另一个版本，我将x和y值缩放了10倍，并旋转到“透视”间隙。如果这是您的选择，您可能需要查看？scaleMatrix