2017-05-27 71 views
1

介绍绘制的校准曲线使用geom_errorbar():格式化数据以GGPLOT2

我有汇总统计的针对三个不同的空气质量测量一个数据帧。仪器名称为aa34aa3548c。它们各自以ppm为单位测量一氧化碳。我有广泛格式的数据,其中每个向量是三种仪器中的每一种的均值,标准差,标准误差或95%置信区间。

我想用绘制和ggplot()这些geom_errorbar()汇总统计,但我有一些麻烦的数据为长格式和ggplot()提供颜色映射的ID变量。我正在关注this教程。下面是我想重现的数字(当然,用有毒气体取代豚鼠的牙齿数据)。我试图添加一个额外的y变量,并让它们通过ID变量进行颜色协调。我期望的输出将用示例中的supp向量替换三个id向量中的两个,即包含aa34aa35的向量。我相当于dose载体将是ref.co.mean,我们的x变量。我相当于len矢量将是长格式的矢量aa34.co.meanaa35.co.mean

link

数据:

## Here's what my data frame looks like. 
## I know it's ugly, but if you copy and paste it into your console it should work! 
df_cal <- structure(list(ref.co.mean = c(1.23638284617457, 1.46466241535712, 
2.16020882959014, 2.55054760052641, 3.13141175081258, 3.86968879644661, 
6.5914211520901), ref.co.sd = c(0.0196205483139859, 0.0229279198586359, 
0.0172965018302434, 0.0164690175286326, 0.00583116470707786, 
0.00975072766851073, 0.0388826652553337), ref.co.se = c(0.00346845569085442, 
0.00193776290206006, 0.00166435666462165, 0.00127061228762621, 
0.000583116470707786, 0.00229826855196908, 0.00614788918523735 
), ref.co.ci = c(0.00707396201972773, 0.00383130164529687, 
0.00329939297398704, 
0.0025085329371034, 0.00115702958592763, 0.00484892279298878, 
0.0124352796323718), id = c("48c", "48c", "48c", "48c", "48c", 
"48c", "48c"), aa34.co.mean = c(0, 0.248857142857143, 0.823777777777778, 
1.256, 1.886, 2.446, 4.54), aa34.co.sd = c(0, 0.0716567783084826, 
0.0660714166547489, 0.0777970497665622, 0.0518459255872629, 0, 
0.0690217357069497), aa34.co.se = c(0, 0.00605610310675521, 
0.0063577250318932, 0.00600217269807407, 0.00518459255872628, 0, 
0.0109132946446067), aa34.co.ci = c(0, 0.0119739921598931, 
0.0126034483753748, 0.0118499152368743, 0.0102873564420935, 0, 
0.0220742219853317), id = c("aa34", "aa34", "aa34", "aa34", "aa34", "aa34", 
"aa34"), aa35.co.mean = c(0.2915625, 0.801035714285714, 1.39911111111111, 
1.80436904761905, 2.45672, 3.02355555555556, 5.134975), aa35.co.sd = 
c(0.0691998633940125, 0.0474980316455754, 0.0846624379229758, 
0.0822798331713915, 0.0595577165445419, 
0.0178768075145867, 0.0243007072942329), aa35.co.se = c(0.0122329231657723, 
0.00401431635364878, 0.00814664688751334, 0.00634802694633388, 
0.00595577165445419, 0.00421360393984362, 0.00384227919014218), aa35.co.ci = 
c(0.0249492112853266, 0.00793701687349159, 0.0161497773125, 
0.0125327252345785, 0.0118175430765459, 0.00888992723110191, 
0.00777174323014678), id = c("aa35", "aa35", "aa35", "aa35", 
"aa35", "aa35", "aa35")), .Names = c("ref.co.mean", "ref.co.sd", 
"ref.co.se", "ref.co.ci", "id", "aa34.co.mean", "aa34.co.sd", 
"aa34.co.se", "aa34.co.ci", "id", "aa35.co.mean", "aa35.co.sd", 
"aa35.co.se", "aa35.co.ci", "id"), row.names = c(1L, 33L, 173L, 
281L, 449L, 549L, 567L), class = "data.frame") 

这是我第一次尝试:

## This code only gets half of the job done... 
## 95% Confidence Intervals for Error Bars: 
p <- ggplot(df_cal, aes(x=ref.co.mean, y=aa34.co.mean)) + 
    theme_bw() + 
    geom_errorbar(aes(ymin=aa34.co.mean-aa34.co.ci, 
         ymax=aa34.co.mean+aa34.co.ci), width =.05) + 
    xlab("Reference CO (ppm)") + 
    ylab("AA34 CO (ppm)") + 
    geom_smooth(method='lm', formula = y~x, se = FALSE) + 
    geom_point(size=2, shape = 21, fill="White") + 
    geom_abline(intercept = 0, slope = 1, color, linetype=2, color = "firebrick") + 
    ggtitle("CO Calibration @ 0% RH") + 
    theme(plot.title = element_text(hjust = 0.5)) + 
    annotate("rect", xmin = 4.80, xmax = 5.70, ymin = 0.70, ymax = 1.70, 
     fill="white", colour="red") + 
    annotate("text", x=5.25, y=1.50, label= "R^2 == 0.994", parse=T) + 
    annotate("text", x=5.25, y=1.20, label= "alpha == -0.9490", parse=T) + 
    annotate("text", x=5.25, y=0.90, label= "beta == 0.849", parse=T) 
p 

link

提前致谢!

+1

您可以编辑您的问题,包括'dput(df_cal)的输出',使这个容易复制? –

+1

此外,我想知道你在哪里计算你的汇总统计?它是“Excel”吗?使用'Rmisc'包中的'SummarySE'函数可以让你更容易,就像你的示例链接一样。 –

+0

@ J.Con感谢'dput()'提示。这不太好,但是如果你复制并粘贴到控制台中,它似乎就可以工作。我仍然使用'R'来计算我的汇总统计。我用'dplyr'从7个时间序列中手动过滤出“高原”。然后,我使用一些基本功能为7个步骤中的每个步骤中的三个仪器中的每一个生成标准偏差,标准误差和95%置信区间的向量。然后,我在7个校准步骤中的每一个上执行'row_bind()',然后是每个校准步骤仅提供一次观察的'unique()'。 – spacedSparking

回答

1

与这里切换到长格式的问题是,你就必须长度7为x轴和y轴2个变量组合长度14的的变量。因此,该解决方案绑定行,以便参考(x轴)数据包含两次。然后在ggplot美学中使用colourgroup

library(ggplot2) 

df_aa34_2<-df_cal[,c(1:4,6:10)]#select first 'aa' group including reference data (48c) 
df_aa35_2<-df_cal[,c(1:4,11:15)]#select second 'aa' group including reference data (48c) 
names(df_aa34_2)<-names(df_aa35_2)#colnames must be the same for rbind function 
DF<-rbind(df_aa34_2,df_aa35_2)#bind rows 

p <- ggplot(DF,aes(x=ref.co.mean,y=aa35.co.mean,colour=id,group=id)) + 
    geom_errorbar(aes(ymin=aa35.co.mean-aa35.co.ci, 
        ymax=aa35.co.mean+aa35.co.ci), width =.5) + 
    xlab("Reference CO (ppm)") + 
    ylab("AA34 & 35 CO (ppm)") + 
    geom_smooth(method='lm', formula = y~x, se = FALSE) + 
    geom_point(size=2, shape = 21, fill="White") + 
    geom_abline(intercept = 0, slope = 1, color, linetype=2, color = "firebrick") + 
    ggtitle("CO Calibration @ 0% RH") + 
    theme(plot.title = element_text(hjust = 0.5)) + 
    annotate("rect", xmin = 4.80, xmax = 5.70, ymin = 0.70, ymax = 1.70, 
      fill="white", colour="red") + 
    annotate("text", x=5.25, y=1.50, label= "R^2 == 0.994", parse=T) + 
    annotate("text", x=5.25, y=1.20, label= "alpha == -0.9490", parse=T) + 
    annotate("text", x=5.25, y=0.90, label= "beta == 0.849", parse=T)+ 
    theme_bw() 
p 

enter image description here

+1

这看起来不错,谢谢!我喜欢你使用'rbind()'而不是尝试使用'reshape2'或'tidyr'。我一直在用'left_join()'将'melt()'函数配对以获得类似格式的数据,但是你的版本只涉及较少的代码行。现在包含一些额外的'annotate()'行来包含第二个'lm()'输出。 – spacedSparking