R中可视化纵向分类数据的好方法

Question

[更新：虽然我已经接受了答案，但如果您有其他可视化想法（无论是R还是其他语言/程序），请添加其他答案。关于分类数据分析的文本似乎没有太多关于可视化纵向数据的内容，而关于纵向数据分析的文本似乎没有太多关于在类别成员资格中随时间变化内部主体变化的可视化。对这个问题有更多的答案将使它成为一个在标准参考文献中没有得到太多报道的问题的更好的资源。]

一位同事刚给我一个纵向分类数据集，我正在试图弄清楚如何捕捉可视化中的纵向方面。我在这里发帖，因为我想在R中这样做，但请告诉我是否有必要交叉发布到Cross-Validated，因为通常不鼓励交叉发布。

快速背景：数据跟踪学生通过学术咨询计划从学期到学期的学术地位。数据为长格式，有五个变量：“id”，“同类”，“术语”，“站立”和“termGPA”。前两个确定学生和他们在咨询计划中的期限。最后三个是学生的学术地位和GPA记录的条款。我使用dput粘贴了下面的一些示例数据。

我创建了一个马赛克图（见下文），通过队列，站立和术语对学生进行分组。这表明在每个学期中，每个学术类别中的学生比例是多少。但这并没有捕捉到纵向方面 - 随着时间的推移追踪个别学生的事实。我想跟踪具有特定学术地位的学生群体随着时间的推移而走的路。

例如：在2009年秋季（“F09”）中具有“AP”（学术试用期）的学生中，在未来的术语中仍然是AP的分数，以及分成其他类别的分数（例如，GS，“良好的信誉”） “）？自进入咨询计划以来，类别之间的移动与时间之间是否存在差异？

我无法弄清楚如何在R图形中捕捉这个纵向方面。 vcd包具有可视化分类数据的功能，但似乎不能解决纵向分类数据。是否存在可视化纵向分类数据的“标准”方法？ R是否有为此设计的包装？长格式适合这种类型的数据，还是宽屏格式会更好？

我希望有解决这一特定问题的建议以及文章，书籍等方面的建议，以便更多地了解纵向分类数据的可视化。

这是我用来制作马赛克图的代码。该代码使用下面列出的数据dput。

library(RColorBrewer)

# create a table object for plotting
df1.tab = table(df1$cohort, df1$term, df1$standing,
            dnn=c("Cohort\nAcademic Standing", "Term", "Standing"))

# create a mosaic plot
plot(df1.tab, las=1, dir=c("h","v","h"), 
     col=brewer.pal(8,"Dark2"),
     main="Fall 2009 and Fall 2010 Cohorts")

这是马赛克图（侧面问题：是否有任何方法可以使F10群组的列直接位于F09群组的下方，并且具有与F09群组相同的宽度，即使F10中的某些术语没有数据也是如此队列）：

以下是用于创建表格和情节的数据：

df1 =
structure(list(id = c(101L, 102L, 103L, 104L, 105L, 106L, 107L, 
108L, 109L, 110L, 111L, 112L, 113L, 114L, 115L, 116L, 117L, 118L, 
119L, 120L, 121L, 122L, 123L, 124L, 125L, 101L, 102L, 103L, 104L, 
105L, 106L, 107L, 108L, 109L, 110L, 111L, 112L, 113L, 114L, 115L, 
116L, 117L, 118L, 119L, 120L, 121L, 122L, 123L, 124L, 125L, 101L, 
102L, 103L, 104L, 105L, 106L, 107L, 108L, 109L, 110L, 111L, 112L, 
113L, 114L, 115L, 116L, 117L, 118L, 119L, 120L, 121L, 122L, 123L, 
124L, 125L, 101L, 102L, 103L, 104L, 105L, 106L, 107L, 108L, 109L, 
110L, 111L, 112L, 113L, 114L, 115L, 116L, 117L, 118L, 119L, 120L, 
121L, 122L, 123L, 124L, 125L, 101L, 102L, 103L, 104L, 105L, 106L, 
107L, 108L, 109L, 110L, 111L, 112L, 113L, 114L, 115L, 116L, 117L, 
118L, 119L, 120L, 121L, 122L, 123L, 124L, 125L, 101L, 102L, 103L, 
104L, 105L, 106L, 107L, 108L, 109L, 110L, 111L, 112L, 113L, 114L, 
115L, 116L, 117L, 118L, 119L, 120L, 121L, 122L, 123L, 124L, 125L, 
101L, 102L, 103L, 104L, 105L, 106L, 107L, 108L, 109L, 110L, 111L, 
112L, 113L, 114L, 115L, 116L, 117L, 118L, 119L, 120L, 121L, 122L, 
123L, 124L, 125L), cohort = structure(c(1L, 1L, 1L, 1L, 2L, 1L, 
1L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 1L, 
1L, 1L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 2L, 2L, 2L, 2L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 1L, 1L, 
2L, 1L, 1L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
1L, 1L, 1L, 1L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 2L, 2L, 2L, 2L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 
1L, 1L, 2L, 1L, 1L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
2L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 2L, 2L, 
2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L, 
1L, 1L, 1L, 1L, 2L, 1L, 1L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L), .Label = c("F09", "F10"), class = c("ordered", 
"factor")), term = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 5L, 
5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 
5L, 5L, 5L, 5L, 5L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 
6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 7L, 7L, 
7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 
7L, 7L, 7L, 7L, 7L, 7L, 7L), .Label = c("S09", "F09", "S10", 
"F10", "S11", "F11", "S12"), class = c("ordered", "factor")), 
    standing = structure(c(2L, 4L, 1L, 4L, NA, 4L, 1L, NA, NA, 
    NA, NA, 2L, 2L, 1L, 4L, 4L, 1L, 3L, NA, NA, 4L, 3L, 1L, 4L, 
    NA, 2L, 1L, 3L, 3L, NA, 1L, 2L, NA, NA, NA, NA, 2L, 4L, 3L, 
    4L, 4L, 4L, 2L, NA, NA, 4L, 2L, 4L, 4L, NA, 3L, 4L, 6L, 6L, 
    1L, 4L, 4L, 1L, 1L, 1L, 1L, 1L, 4L, 6L, 4L, 4L, 1L, 4L, 1L, 
    2L, 4L, 3L, 1L, 4L, 1L, 6L, 1L, 6L, 6L, 7L, 4L, 4L, 2L, 2L, 
    4L, 2L, 6L, 4L, 6L, 7L, 4L, 2L, 4L, 1L, 2L, 4L, 6L, 6L, 4L, 
    2L, 2L, 3L, 6L, 6L, 7L, 4L, 4L, 3L, 4L, 4L, 6L, 2L, 1L, 6L, 
    6L, 4L, 2L, 1L, 7L, 2L, 4L, 6L, 6L, 4L, 4L, 3L, 6L, 4L, 6L, 
    2L, 4L, 4L, 6L, 4L, 4L, 6L, 3L, 2L, 6L, 6L, 4L, 2L, 6L, 3L, 
    4L, 4L, 6L, 6L, 4L, 4L, 5L, 6L, 4L, 6L, 4L, 4L, 4L, 5L, 4L, 
    4L, 6L, 6L, 2L, 6L, 6L, 4L, 3L, 6L, 6L, 4L, 4L, 6L, 6L, 4L, 
    4L), .Label = c("AP", "CP", "DQ", "GS", "DM", "NE", "WD"), class = "factor"), 
    termGPA = c(1.433, 1.925, 1, 1.68, NA, 1.579, 1.233, NA, 
    NA, NA, NA, 2.009, 1.675, 0, 1.5, 1.86, 0.5, 0.94, NA, NA, 
    1.777, 1.1, 1.133, 1.675, NA, 2, 1.25, 1.66, 0, NA, 1.525, 
    2.25, NA, NA, NA, NA, 1.66, 2.325, 0, 2.308, 1.6, 1.825, 
    2.33, NA, NA, 2.65, 2.65, 2.85, 3.233, NA, 1.25, 1.575, NA, 
    NA, 1, 2.385, 3.133, 0, 0, 1.729, 1.075, 0, 4, NA, 2.74, 
    0, 1.369, 2.53, 0, 2.65, 2.75, 0, 0.333, 3.367, 1, NA, 0.1, 
    NA, NA, 1, 2.2, 2.18, 2.31, 1.75, 3.073, 0.7, NA, 1.425, 
    NA, 2.74, 2.9, 0.692, 2, 0.75, 1.675, 2.4, NA, NA, 3.829, 
    2.33, 2.3, 1.5, NA, NA, NA, 2.69, 1.52, 0.838, 2.35, 1.55, 
    NA, 1.35, 0.66, NA, NA, 1.35, 1.9, 1.04, NA, 1.464, 2.94, 
    NA, NA, 3.72, 2.867, 1.467, NA, 3.133, NA, 1, 2.458, 1.214, 
    NA, 3.325, 2.315, NA, 1, 2.233, NA, NA, 2.567, 1, NA, 0, 
    3.325, 2.077, NA, NA, 3.85, 2.718, 1.385, NA, 2.333, NA, 
    2.675, 1.267, 1.6, 1.388, 3.433, 0.838, NA, NA, 0, NA, NA, 
    2.6, 0, NA, NA, 1, 2.825, NA, NA, 3.838, 2.883)), .Names = c("id", 
"cohort", "term", "standing", "termGPA"), row.names = c("101.F09.s09", 
"102.F09.s09", "103.F09.s09", "104.F09.s09", "105.F10.s09", "106.F09.s09", 
"107.F09.s09", "108.F10.s09", "109.F10.s09", "110.F10.s09", "111.F10.s09", 
"112.F09.s09", "113.F09.s09", "114.F09.s09", "115.F09.s09", "116.F09.s09", 
"117.F09.s09", "118.F09.s09", "119.F10.s09", "120.F10.s09", "121.F09.s09", 
"122.F09.s09", "123.F09.s09", "124.F09.s09", "125.F10.s09", "101.F09.f09", 
"102.F09.f09", "103.F09.f09", "104.F09.f09", "105.F10.f09", "106.F09.f09", 
"107.F09.f09", "108.F10.f09", "109.F10.f09", "110.F10.f09", "111.F10.f09", 
"112.F09.f09", "113.F09.f09", "114.F09.f09", "115.F09.f09", "116.F09.f09", 
"117.F09.f09", "118.F09.f09", "119.F10.f09", "120.F10.f09", "121.F09.f09", 
"122.F09.f09", "123.F09.f09", "124.F09.f09", "125.F10.f09", "101.F09.s10", 
"102.F09.s10", "103.F09.s10", "104.F09.s10", "105.F10.s10", "106.F09.s10", 
"107.F09.s10", "108.F10.s10", "109.F10.s10", "110.F10.s10", "111.F10.s10", 
"112.F09.s10", "113.F09.s10", "114.F09.s10", "115.F09.s10", "116.F09.s10", 
"117.F09.s10", "118.F09.s10", "119.F10.s10", "120.F10.s10", "121.F09.s10", 
"122.F09.s10", "123.F09.s10", "124.F09.s10", "125.F10.s10", "101.F09.f10", 
"102.F09.f10", "103.F09.f10", "104.F09.f10", "105.F10.f10", "106.F09.f10", 
"107.F09.f10", "108.F10.f10", "109.F10.f10", "110.F10.f10", "111.F10.f10", 
"112.F09.f10", "113.F09.f10", "114.F09.f10", "115.F09.f10", "116.F09.f10", 
"117.F09.f10", "118.F09.f10", "119.F10.f10", "120.F10.f10", "121.F09.f10", 
"122.F09.f10", "123.F09.f10", "124.F09.f10", "125.F10.f10", "101.F09.s11", 
"102.F09.s11", "103.F09.s11", "104.F09.s11", "105.F10.s11", "106.F09.s11", 
"107.F09.s11", "108.F10.s11", "109.F10.s11", "110.F10.s11", "111.F10.s11", 
"112.F09.s11", "113.F09.s11", "114.F09.s11", "115.F09.s11", "116.F09.s11", 
"117.F09.s11", "118.F09.s11", "119.F10.s11", "120.F10.s11", "121.F09.s11", 
"122.F09.s11", "123.F09.s11", "124.F09.s11", "125.F10.s11", "101.F09.f11", 
"102.F09.f11", "103.F09.f11", "104.F09.f11", "105.F10.f11", "106.F09.f11", 
"107.F09.f11", "108.F10.f11", "109.F10.f11", "110.F10.f11", "111.F10.f11", 
"112.F09.f11", "113.F09.f11", "114.F09.f11", "115.F09.f11", "116.F09.f11", 
"117.F09.f11", "118.F09.f11", "119.F10.f11", "120.F10.f11", "121.F09.f11", 
"122.F09.f11", "123.F09.f11", "124.F09.f11", "125.F10.f11", "101.F09.s12", 
"102.F09.s12", "103.F09.s12", "104.F09.s12", "105.F10.s12", "106.F09.s12", 
"107.F09.s12", "108.F10.s12", "109.F10.s12", "110.F10.s12", "111.F10.s12", 
"112.F09.s12", "113.F09.s12", "114.F09.s12", "115.F09.s12", "116.F09.s12", 
"117.F09.s12", "118.F09.s12", "119.F10.s12", "120.F10.s12", "121.F09.s12", 
"122.F09.s12", "123.F09.s12", "124.F09.s12", "125.F10.s12"), reshapeLong = structure(list(
    varying = list(c("s09as", "f09as", "s10as", "f10as", "s11as", 
    "f11as", "s12as"), c("s09termGPA", "f09termGPA", "s10termGPA", 
    "f10termGPA", "s11termGPA", "f11termGPA", "s12termGPA")), 
    v.names = c("standing", "termGPA"), idvar = c("id", "cohort"
    ), timevar = "term"), .Names = c("varying", "v.names", "idvar", 
"timevar")), class = "data.frame")

Answer 1

以下是绘制数据的一些想法。我已经使用了ggplot2，并且我已经在某些地方重新格式化了数据。

图1

我用堆叠的条形图来模仿你的马赛克图并解决对齐问题。

图2

每个学生的数据点用灰线连接，这使得人们想起平行坐标图。着色点显示了分类的立场。在y轴上使用GPA有助于扩展点以减少过度绘图，并显示站立和GPA的相关性。一个主要问题是许多有效standing数据点丢失，因为它们缺少匹配的termGPA值。

图3

在这里，我创建了一个名为initial_standing的新变量，用于facetting。每个小组都包含在同一队列和initial_standing中匹配的学生。将id绘制为文本会使这个数字变得混乱，但在某些情况下可能会有用。

图4

这个情节就像一个热图，每一行都是学生。我控制了id轴的顺序，迫使initial_standing和同类群组保持在一起。如果您有更多行，则可能需要考虑按某种类型的聚类对行进行排序。

library(ggplot2)

# Create new data frame for determining initial standing.
standing_data = data.frame(id=unique(df1$id), initial_standing=NA, cohort=NA)

for (i in 1:nrow(standing_data)) {
    id = standing_data$id[i]
    subdat = df1[df1$id == id, ]
    subdat = subdat[complete.cases(subdat), ]
    initial_standing = subdat$standing[which.min(subdat$term)]
    standing_data[i, "initial_standing"] = as.character(initial_standing)
    standing_data[i, "cohort"] = as.character(subdat$cohort[1])
}

standing_data$cohort = factor(standing_data$cohort, levels=levels(df1$cohort))
standing_data$initial_standing = factor(standing_data$initial_standing,
                                        levels=levels(df1$standing))

# Add the new column (initial_standing) to df1.
df1 = merge(df1, standing_data[, c("id", "initial_standing")], by="id")

# Remove rows where standing is missing. Make some plots tidier.
df1 = df1[!is.na(df1$standing), ]

# Create id factor, controlling the sort order of the levels.     
id_order = order(standing_data$initial_standing, standing_data$cohort)
df1$id = factor(df1$id, levels=as.character(standing_data$id)[id_order])


p1 = ggplot(df1, aes(x=term, fill=standing)) +
     geom_bar(position="fill", colour="grey20", size=0.5, width=1.0) +
     facet_grid(cohort ~ .) +
     scale_fill_brewer(palette="Set1")

p2 = ggplot(df1, aes(x=term, y=termGPA, group=id)) + 
     geom_line(colour="grey70") + 
     geom_point(aes(colour=standing), size=4) + 
     facet_grid(cohort ~ .) +
     scale_colour_brewer(palette="Set1")

p3 = ggplot(df1, aes(x=term, y=termGPA, group=id)) +
     geom_line(colour="grey70") + 
     geom_point(aes(colour=standing), size=4) + 
     geom_text(aes(label=id), hjust=-0.30, size=3) +
     facet_grid(initial_standing ~ cohort) +
     scale_colour_brewer(palette="Set1")


p4 = ggplot(df1, aes(x=term, y=id, fill=standing)) + 
     geom_tile(colour="grey20") +
     facet_grid(initial_standing ~ ., space="free_y", scales="free_y") +
     scale_fill_brewer(palette="Set1") +
     opts(panel.grid.major=theme_blank()) +
     opts(panel.grid.minor=theme_blank())

ggsave("plot_1.png", p1, width=10, height=6.25, dpi=80)
ggsave("plot_2.png", p2, width=10, height=6.25, dpi=80)
ggsave("plot_3.png", p3, width=10, height=6.25, dpi=80)
ggsave("plot_4.png", p4, width=10, height=6.25, dpi=80)

Answer 2

在研究我的问题时，我发现了一些其他选项，我将在这里列出。

许多相对较新的R包被设计用于可视化和分析“生活史”或“多状态序列”数据。这个想法是随着时间的推移，人们（或物体）进入和退出各种类别 - 例如，职业变化，婚姻和离婚，健康和疾病，或者在我的情况下，大学的学术地位类别。

用于可视化序列或生活历史数据的R包包括biograph，由@timriffe在上面的评论中提及，以及TraMineR。传记包的作者弗兰斯威勒肯斯（Frans Willekens）有一本关于包装的书，传记。 R 对生活史进行多态分析，将于今年秋季由Springer出版。 TraMineR在上面的链接中有详细的用户手册，还有一个较短的JSS article。 JSS还有一个special issue on multi-state models in the context of risk analysis，讨论了用于多态建模的其他R包。

我还发现了一些专门的软件，用于可视化类别之间的移动。 Parallel Sets是一个简单的免费程序，用于生成基本可视化，但它的灵活性有限。 Lifeflow更复杂。它也是免费的，但您必须向创建者发送电子邮件，要求提供副本。

一旦我有机会试用这些工具，我会在这个答案中添加更多细节。

Answer 3

我希望在写一个R软件包来解决这个问题之前我找到了@bdemarest的答案，但是由于OP请求了额外的更新，我将再分享一个解决方案。 bdemarest在图4中提出的是我所谓的一种水平线图。

在开发longCatEDA R软件包时，我们发现对数据进行排序对于制作有用的图表至关重要（请参阅example(sorter)以及下面评论中链接的报告以获取技术细节），尤其是尺寸问题变得很大。例如，我们在3年（> 1000天）内为数千名参与者提供了每日饮酒数据（禁用，使用，滥用）的问题。

将水平线图应用于@ eipi10数据的代码如下。图1按term分层，图2按照第一状态分层，如图4中的@bdemarest，尽管由于分层分类，结果并不相同。

图1

图2

# libraries
install.packages('longCatEDA')
library(longCatEDA)
library(RColorBrewer)

# transform data long to wide
dfw <- reshape(df1,
           timevar = 'term',
           idvar = c('id', 'cohort'),
           direction = 'wide')

# set up objects required by longCat()
y <- dfw[,seq(3,15,by=2)]
Labels <- levels(df1$standing)
tLabels <- levels(df1$term)
groupLabels <- levels(dfw$cohort)

# use the same colors as bdemarest
cols <- brewer.pal(7, "Set1")

# plot the longCat object
png('plot1.png', width=10, height=6.25, units='in', res=100)
par(bg='cornsilk3', mar=c(5.1, 4.1, 4.1, 8.1), xpd=TRUE)
lc <- longCat(y=y, Labels=Labels, tLabels=tLabels, id=dfw$id) 
longCatPlot(lc, cols=cols, xlab='Term', lwd=8, legendBuffer=0)
legend(8.1, 25, legend=Labels, col=cols, lty=1, lwd=4)
dev.off()

# stratify by term
png('plot2.png', width=10, height=6.25, units='in', res=100)
par(bg='cornsilk3', mar=c(5.1, 4.1, 4.1, 8.1), xpd=TRUE)
lc.g <- sorter(lc, group=dfw$cohort, groupLabels=groupLabels)
longCatPlot(lc.g, cols=cols, xlab='Term', lwd=8, legendBuffer=0) 
legend(8.1, 25, legend=Labels, col=cols, lty=1, lwd=4)
dev.off()

# stratify by first status, akin to Figure 4 by bdemarest
png('plot2.png', width=10, height=6.25, units='in', res=100)
par(bg='cornsilk3', mar=c(5.1, 4.1, 4.1, 8.1), xpd=TRUE)
first <- apply(!is.na(y), 1, function(x) which(x)[1])
first <- y[cbind(seq_along(first), first)]
lc.1 <- sorter(lc, group=factor(first), groupLabels = sort(unique(first)))
longCatPlot(lc.1, cols=cols, xlab='Term', lwd=8, legendBuffer=0) 
legend(8.1, 25, legend=Labels, col=cols, lty=1, lwd=4)
dev.off()