R 使用镶嵌面“U形包裹和比例=”设置单个轴限制;“免费”;在ggplot2中

R 使用镶嵌面“U形包裹和比例=”设置单个轴限制;“免费”;在ggplot2中,r,ggplot2,facet,R,Ggplot2,Facet,我正在创建一个分面图,用预测值与残差图并排查看预测值与实际值。我将使用shinny帮助探索使用不同训练参数进行建模的结果。我用85%的数据训练模型,对剩余的15%进行测试,并重复这5次,每次收集实际/预测值。计算残差后,我的data.frame如下所示: head(results) act pred resid 2 52.81000 52.86750 -0.05750133 3 44.46000 42.76825 1.69175252 4 54.58667 4

我正在创建一个分面图,用预测值与残差图并排查看预测值与实际值。我将使用
shinny
帮助探索使用不同训练参数进行建模的结果。我用85%的数据训练模型,对剩余的15%进行测试,并重复这5次,每次收集实际/预测值。计算残差后,我的
data.frame
如下所示:

head(results)
       act     pred       resid
2 52.81000 52.86750 -0.05750133
3 44.46000 42.76825  1.69175252
4 54.58667 49.00482  5.58184181
5 36.23333 35.52386  0.70947731
6 53.22667 48.79429  4.43237981
7 41.72333 41.57504  0.14829173
我想要的是:

  • pred
    act
    pred
    resid
  • pred
    act
    的x/y范围/限制相同,理想情况下从
    min(min(results$act)、min(results$pred))
    max(max(results$act)、max(results$pred))
  • pred
    resid
    的x/y范围/限制不受我对实际与预测图所做的操作的影响。仅在预测值上绘制
    x
    ,仅在剩余范围上绘制
    y
    ,即可
为了并排查看两个图,我融合了数据:

library(reshape2)
plot <- melt(results, id.vars = "pred")
但这会拾取剩余值的
min()

我的最后一个想法是在熔化之前存储最小
act
pred
变量的值,然后将它们添加到熔化的数据帧中,以便指示它们出现在哪个方面:

head(results)
       act     pred       resid
2 52.81000 52.86750 -0.05750133
3 44.46000 42.76825  1.69175252
4 54.58667 49.00482  5.58184181
5 36.23333 35.52386  0.70947731

min_xy <- min(min(results$act), min(results$pred))
max_xy <- max(max(results$act), max(results$pred))

plot <- melt(results, id.vars = "pred")

plot <- rbind(plot, data.frame(pred = c(min_xy, max_xy),
  variable = c("act", "act"), value = c(max_xy, min_xy)))

p <- ggplot(plot, aes(x = pred, y = value)) + geom_point(size = 2.5) + theme_bw()
p <- p + facet_wrap(~variable, scales = "free")

print(p)

我不确定我是否理解你想要什么,但基于我所理解的

x比例似乎相同,但y比例不同,这是因为您指定了scales=“free”

您可以指定scales=“free_x”以仅允许x自由(在这种情况下,它与pred定义的范围相同)


p这是一些带有虚拟
geom_blank
层的代码

range_act <- range(range(results$act), range(results$pred))

d <- reshape2::melt(results, id.vars = "pred")

dummy <- data.frame(pred = range_act, value = range_act,
                    variable = "act", stringsAsFactors=FALSE)

ggplot(d, aes(x = pred, y = value)) +
  facet_wrap(~variable, scales = "free") +
  geom_point(size = 2.5) + 
  geom_blank(data=dummy) + 
  theme_bw()

range\u act您还可以使用坐标笛卡尔命令指定范围,以设置所需的y轴范围,如前一个使用后比例=自由x

p <- ggplot(plot, aes(x = pred, y = value)) +
     geom_point(size = 2.5) +
     theme_bw()+
     coord_cartesian(ylim = c(-20, 80))
p <- p + facet_wrap(~variable, scales = "free_x")
p

p只是好奇-为什么不在同一个图形中绘制实际和残差?我会分别创建这些图,然后使用
网格。排列
@RicardoSaporta有没有可以链接到的谷歌图像作为示例?您是否建议,使用融化后的数据,我只做
ggplot(plot,aes(x=pred,y=value))+geom_point()
,而不做刻面?这真的会缩小残差的规模,使其难以检测非随机性/偏斜吗?我的另一个评论是,刻面更少的代码。。。我只需要熔化,然后通过
melt()
创建的
变量
值来绘制和刻面。然后,我想我可以将它们存储在由
lappy
创建的列表中,以绘制各种组合。谢谢你的意见。如果你想创建一个
网格
解决方案,我可以接受这个答案,不过如果这是我们采取的路线,这也可能是其他基于
网格
的解决方案的重复。@joran和我发现我经常建议人们不要使用
网格。整理
,这几乎总是会弄乱布局。我希望GTTable的长期缺陷得到解决。一个很好的变体是
expand\u limits(pred=range\u act,value=range\u act)
,它使用
geom\u blank
,但使用起来更简单。这只会扩展限制(但不会收缩限制)。有办法缩短范围吗@巴蒂斯特
> dput(results)
structure(list(act = c(52.81, 44.46, 54.5866666666667, 36.2333333333333, 
53.2266666666667, 41.7233333333333, 35.2966666666667, 30.6833333333333, 
39.25, 35.8866666666667, 25.1, 29.0466666666667, 23.2766666666667, 
56.3866666666667, 42.92, 41.57, 27.92, 23.16, 38.0166666666667, 
61.8966666666667, 37.41, 41.6333333333333, 35.9466666666667, 
48.9933333333333, 30.5666666666667, 32.08, 40.3633333333333, 
53.2266666666667, 64.6066666666667, 38.5366666666667, 41.7233333333333, 
25.78, 33.4066666666667, 27.8033333333333, 39.3266666666667, 
48.9933333333333, 25.2433333333333, 32.67, 55.17, 42.92, 54.5866666666667, 
23.16, 64.6066666666667, 40.7966666666667, 39.0166666666667, 
41.6333333333333, 35.8866666666667, 25.1, 23.2766666666667, 44.46, 
34.2166666666667, 40.8033333333333, 24.5766666666667, 35.73, 
61.8966666666667, 62.1833333333333, 74.6466666666667, 39.4366666666667, 
36.6, 27.1333333333333), pred = c(52.8675013282404, 42.7682474758679, 
49.0048248585123, 35.5238560262515, 48.7942868566949, 41.5750416040131, 
33.9548164913007, 29.9787449128663, 37.6443975781139, 36.7196211666685, 
27.6043278172077, 27.0615724310721, 31.2073056885252, 55.0886903524179, 
43.0895814712768, 43.0895814712768, 32.3549865881578, 26.2428426737583, 
36.6926037128343, 56.7987490221996, 45.0370788180147, 41.8231642271826, 
38.3297859332601, 49.5343916620086, 30.8535641206809, 29.0117492750411, 
36.9767968381391, 49.0826677983065, 54.4678549541069, 35.5059204731218, 
41.5333417555995, 27.6069075391361, 31.2404889715121, 27.8920960978598, 
37.8505531149324, 49.2616631533957, 30.366837650159, 31.1623492639066, 
55.0456078770405, 42.772538591063, 49.2419293590535, 26.1963523976241, 
54.4080781796616, 44.9796700541254, 34.6996927469131, 41.6227713664027, 
36.8449646519306, 27.5318686661673, 31.6641793552795, 42.8198894266632, 
40.5769177148146, 40.5769177148146, 29.3807781312816, 36.8579132935989, 
55.5617033901752, 55.8097119335638, 55.1041728261666, 43.6094641699075, 
37.0674887276681, 27.3876960746536), resid = c(-0.0575013282403773, 
1.69175252413213, 5.58184180815435, 0.709477307081826, 4.43237980997177, 
0.148291729320228, 1.34185017536599, 0.704588420467079, 1.60560242188613, 
-0.832954500001826, -2.50432781720766, 1.98509423559461, -7.93063902185855, 
1.29797631424874, -0.169581471276786, -1.51958147127679, -4.43498658815778, 
-3.08284267375831, 1.32406295383237, 5.09791764446704, -7.62707881801468, 
-0.189830893849219, -2.38311926659339, -0.541058328675241, -0.286897454014273, 
3.06825072495888, 3.38653649519422, 4.14399886836018, 10.1388117125598, 
3.03074619354486, 0.189991577733821, -1.82690753913609, 2.16617769515461, 
-0.088762764526507, 1.47611355173427, -0.268329820062384, -5.12350431682565, 
1.5076507360934, 0.124392122959534, 0.147461408936991, 5.34473730761318, 
-3.03635239762411, 10.1985884870051, -4.18300338745873, 4.31697391975358, 
0.0105619669306023, -0.958297985263961, -2.43186866616734, -8.38751268861282, 
1.64011057333683, -6.36025104814794, 0.226415618518729, -4.80411146461488, 
-1.1279132935989, 6.33496327649151, 6.37362139976954, 19.5424938405001, 
-4.17279750324084, -0.467488727668119, -0.254362741320246)), .Names = c("act", 
"pred", "resid"), row.names = c(2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 
10L, 11L, 12L, 13L, 15L, 16L, 17L, 18L, 19L, 20L, 21L, 22L, 23L, 
24L, 25L, 26L, 28L, 29L, 30L, 31L, 32L, 33L, 34L, 35L, 36L, 37L, 
38L, 39L, 41L, 42L, 43L, 44L, 45L, 46L, 47L, 48L, 49L, 50L, 51L, 
52L, 54L, 55L, 56L, 57L, 58L, 59L, 60L, 61L, 62L, 63L, 64L, 65L
), class = "data.frame")
p <- ggplot(plot, aes(x = pred, y = value)) + geom_point(size = 2.5) + theme_bw()
p <- p + facet_wrap(~variable, scales = "free_x")
range_act <- range(range(results$act), range(results$pred))

d <- reshape2::melt(results, id.vars = "pred")

dummy <- data.frame(pred = range_act, value = range_act,
                    variable = "act", stringsAsFactors=FALSE)

ggplot(d, aes(x = pred, y = value)) +
  facet_wrap(~variable, scales = "free") +
  geom_point(size = 2.5) + 
  geom_blank(data=dummy) + 
  theme_bw()
p <- ggplot(plot, aes(x = pred, y = value)) +
     geom_point(size = 2.5) +
     theme_bw()+
     coord_cartesian(ylim = c(-20, 80))
p <- p + facet_wrap(~variable, scales = "free_x")
p