如何在R语言中反转广义加性模型3D透视图的Z轴？

问题描述 #

我绘制了广义加性模型（generalized additive models）的3D透视图，但Z轴Growers (proportion)的数值默认是从上到下读取的，希望反转Z轴，改为从下到上显示数值。

可复现代码：

n <- 200
sig2 <- 4
x1 <- runif(n, 0, 1)
x2 <- runif(n, 0, 1)
x3 <- runif(n, 0, 1)
x4 <- runif(n, 0, 1)
y <- x1^2 + x1 * x2 * x4

mod <- gam(y ~ s(x1, x2) + s(x3, x4))


p1 <- wrap_elements(panel = ~vis.gam(mod, view = c("x1",  "x2"), zlab = "Growers (proportion)", 
                                     theta = 120, plot.type = "persp", type = "response", ticktype = "detailed"))
p2 <- wrap_elements(panel = ~vis.gam(mod, view = c("x3",  "x4"), zlab = "Growers (proportion)", 
                                     theta = 35, plot.type = "persp", type = "response", ticktype = "detailed"))

fig <- p1 + p2 + plot_layout(ncol = 2, tag_level = 'new') +
  plot_annotation(tag_levels = list(c('(a)', '(b)'))) 
fig

解决方案 #

要反转vis.gam生成的3D透视图Z轴，核心是通过反转Z轴的数值范围实现方向调整，以下是两种可行方法：

方法1：自动获取并反转Z轴范围 #

先计算模型预测值的范围，再通过rev()反转后传入zlim参数，确保轴范围准确匹配数据：

# 计算x1-x2维度的Z值范围并反转
z_range1 <- range(predict(mod, newdata = expand.grid(x1=seq(0,1,length=100), x2=seq(0,1,length=100))))
p1 <- wrap_elements(panel = ~vis.gam(mod, view = c("x1",  "x2"), zlab = "Growers (proportion)", 
                                     theta = 120, plot.type = "persp", type = "response", ticktype = "detailed",
                                     zlim = rev(z_range1)))

# 计算x3-x4维度的Z值范围并反转
z_range2 <- range(predict(mod, newdata = expand.grid(x3=seq(0,1,length=100), x4=seq(0,1,length=100))))
p2 <- wrap_elements(panel = ~vis.gam(mod, view = c("x3",  "x4"), zlab = "Growers (proportion)", 
                                     theta = 35, plot.type = "persp", type = "response", ticktype = "detailed",
                                     zlim = rev(z_range2)))

fig <- p1 + p2 + plot_layout(ncol = 2, tag_level = 'new') +
  plot_annotation(tag_levels = list(c('(a)', '(b)'))) 
fig

方法2：手动指定反转后的Z轴范围 #

如果已知Z值的大致区间（比如0到1），可以直接手动传入反转后的范围：

p1 <- wrap_elements(panel = ~vis.gam(mod, view = c("x1",  "x2"), zlab = "Growers (proportion)", 
                                     theta = 120, plot.type = "persp", type = "response", ticktype = "detailed",
                                     zlim = c(1, 0)))  # 手动指定反转范围
p2 <- wrap_elements(panel = ~vis.gam(mod, view = c("x3",  "x4"), zlab = "Growers (proportion)", 
                                     theta = 35, plot.type = "persp", type = "response", ticktype = "detailed",
                                     zlim = c(1, 0)))

fig <- p1 + p2 + plot_layout(ncol = 2, tag_level = 'new') +
  plot_annotation(tag_levels = list(c('(a)', '(b)'))) 
fig

原理说明 #

vis.gam的plot.type="persp"本质调用基础绘图的persp函数，该函数通过zlim参数控制Z轴范围。当范围的最大值在前、最小值在后时，轴的显示方向会自动反转，从而实现数值从下到上递增的效果。

内容的提问来源于stack exchange，提问作者Ahsk