ggparcoord平行坐标图问题：数据显示不全与坐标轴顺序调整

我用ggparcoord处理含152条观测的数据集，绘制平行坐标图，X轴指定为MCA标准化维度变量Std_Dim1至Std_Dim4，按cluster变量分组，遇到两个问题：

• 图中显示的点数远少于152，怀疑数据未全部展示，同时收到警告：In summary.lm(lm(x ~ as.factor(classVar == class.names[i]))) : essentially perfect fit: summary may be unreliable，无法理解含义；
• 图中Std_Dim4排在X轴首位，希望按Std_Dim1到Std_Dim4的顺序排列坐标轴。

附数据片段及原代码：

# 数据片段
structure(list(ID = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 
13, 14, 15, 16, 17, 18, 19, 20), cluster = structure(c(1L, 2L, 
2L, 2L, 2L, 1L, 3L, 1L, 2L, 2L, 3L, 2L, 1L, 4L, 2L, 1L, 5L, 1L, 
3L, 5L), levels = c("1", "2", "3", "4", "5", "6"), class = "factor"), 
    Std_Dim1 = c(-0.0728923703380469, -0.339155028394344, 0.340473420392532, 
    -0.339155028394344, -0.339155028394344, -0.0728923703380469, 
    -0.267300038214234, -0.0728923703380469, -0.339155028394344, 
    -0.339155028394344, -0.267300038214234, -0.339155028394344, 
    -0.0728923703380469, 8.24372395142489, -0.339155028394344, 
    -0.0728923703380469, -0.340751805953902, 1.23367086777029, 
    -0.144946957713102, 0.895154025143995), Std_Dim2 = c(0.380193300283392, 
    -0.21657689529506, -0.0605737618430261, -0.21657689529506, 
    -0.21657689529506, 0.380193300283392, 0.0358222042004695, 
    0.380193300283392, -0.21657689529506, -0.21657689529506, 
    0.0358222042004695, -0.21657689529506, 0.380193300283392, 
    2.24175307931177, -0.21657689529506, 0.380193300283392, 4.41531912524421, 
    0.549235501606044, 0.127794200787863, 1.07941331484527), 
    Std_Dim3 = c(-0.249302376419046, -0.301241388482618, -0.515922638345379, 
    -0.301241388482618, -0.301241388482618, -0.249302376419046, 
    -0.243613817954941, -0.249302376419046, -0.301241388482618, 
    -0.301241388482618, -0.243613817954941, -0.301241388482618, 
    -0.249302376419046, -2.55242656849512, -0.301241388482618, 
    -0.249302376419046, 3.97061868943169, -0.505287507303791, 
    -0.306929946946723, 0.594582905299551), Std_Dim4 = c(-1.78310815898572, 
    0.600501349348233, 0.970728782381305, 0.600501349348233, 
    0.600501349348233, -1.78310815898572, -0.595475416793659, 
    -1.78310815898572, 0.600501349348233, 0.600501349348233, 
    -0.595475416793659, 0.600501349348233, -1.78310815898572, 
    2.19760934092996, 0.600501349348233, -1.78310815898572, 0.0136383315437233, 
    -1.23858333677849, -0.586822354919764, 1.81841980809893)), row.names = c(NA, 
-20L), class = c("tbl_df", "tbl", "data.frame"))

# 原代码
ggparcoord(res.cluster,
           columns = 21:24, groupColumn = 2, order = "anyClass",
           showPoints = TRUE, 
           title = "Location of cluster points in the MCA dimensions",
           alphaLines = 0.5
) + 
  scale_color_viridis(discrete=TRUE) +
  theme_ipsum()+
  theme(
    plot.title = element_text(size=10)
  )+
  theme(text=element_text(family="Calibri"))+
  xlab("")

问题1：点数缺失与警告解释 #

警告含义 #

该警告来自ggparcoord内部的坐标轴排序逻辑：当设置order = "anyClass"时，函数会为每个类别拟合线性模型来确定坐标轴顺序。如果某个维度下，同一类别的观测值完全一致（比如数据片段中cluster 1的所有Std_Dim1值相同），线性模型会出现完全拟合，导致统计量失去意义，因此抛出该警告。

点数缺失原因 #

视觉上点数少是因为大量观测点重合——同一cluster的多条观测在所有维度上数值完全一致，绘图时重叠在一起，看起来像一个点。可以通过调整点的尺寸和透明度让重叠点更明显：

showPoints = TRUE, pointSize = 2, pointAlpha = 0.3

若不需要坐标轴自动排序，可将order设为"none"，既能消除警告，也能保留原始变量顺序：

order = "none"

问题2：坐标轴顺序调整 #

核心解决方法 #

原代码用列索引21:24指定X轴变量，若数据集列顺序变动或索引对应错误，会导致坐标轴混乱。直接用变量名指定，同时配合order = "none"强制保留指定顺序：

columns = c("Std_Dim1", "Std_Dim2", "Std_Dim3", "Std_Dim4")
order = "none"

修改后的完整代码 #

ggparcoord(res.cluster,
           columns = c("Std_Dim1", "Std_Dim2", "Std_Dim3", "Std_Dim4"), 
           groupColumn = "cluster", 
           order = "none",
           showPoints = TRUE, 
           pointSize = 2,
           pointAlpha = 0.3,
           title = "Location of cluster points in the MCA dimensions",
           alphaLines = 0.5
) + 
  scale_color_viridis(discrete=TRUE) +
  theme_ipsum()+
  theme(
    plot.title = element_text(size=10),
    text=element_text(family="Calibri")
  )+
  xlab("")

内容的提问来源于stack exchange，提问作者elsich