I want ggplot() to label observations with residuals higher than 1.5 times the standard error of the regression. The data are these (from Frank 1984):
d <- data.frame(x=c(43,32,32,30,26,25,23,22,22,21,20,20,19,19,19,18,18,17,17,16,16,16,15,13,12,12,10,10,9,7,6,3), y=c(63.0,54.3,51.0,39.0,52.0,55.0,41.2,47.7,44.5,43.0,46.8,42.4,56.5,55.0,53.0,55.0,45.0,50.7,37.5,61.0,48.1,30.0,51.5,40.6,51.3,50.3,62.4,39.3,43.2,40.4,37.7,27.7))
The model is simple:
m <- lm(y~x,data=d)
Then the ggplot() is:
ggplot(d, aes(x=x, y=y)) + geom_point() + geom_text(label=ifelse(abs(resid(m))>(1.5*sigma(m)),rownames(d),""),
nudge_x = 1, nudge_y = 0, check_overlap = T, color="blue")
giving this plot
which is missing a label for the observation in the top left corner (obs #27). Compare:
abs(resid(m))>(1.5*sigma(m))
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
FALSE FALSE FALSE TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE TRUE FALSE FALSE FALSE FALSE TRUE FALSE FALSE FALSE FALSE TRUE
which indicates correctly that 27 satisfies the condition. Why is it not labelled?
Your labels in geom_text aren't inside an aes like they should be, although I'm unsure why you still got partially working labels without it.
I'm including some intermediate steps to work through this more slowly; for me, that helps with debugging and investigating how things work. Feel free to condense.
Assigning d and m are identical to the OP. With the extra steps:
library(tidyverse)
d2 <- d %>%
mutate(row = row_number()) %>%
mutate(abs_resid = abs(resid(m)), sig = sigma(m)) %>%
mutate(is_outlier = abs_resid > 1.5 * sig) %>%
mutate(label = ifelse(is_outlier, row, ""))
head(d2)
#> x y row abs_resid sig is_outlier label
#> 1 43 63.0 1 4.8398378 7.934235 FALSE
#> 2 32 54.3 2 0.9561793 7.934235 FALSE
#> 3 32 51.0 3 2.3438207 7.934235 FALSE
#> 4 30 39.0 4 13.4681223 7.934235 TRUE 4
#> 5 26 52.0 5 1.2832746 7.934235 FALSE
#> 6 25 55.0 6 4.7211239 7.934235 FALSE
ggplot(d2, aes(x = x, y = y)) +
geom_point() +
geom_text(aes(label = label), nudge_x = 1, color = "blue")
Created on 2018-07-31 by the reprex package (v0.2.0).
Related
I have a series of data of 60,000 data which part of the data is as the figure 1 (the whole curve is not so nice and uniform like this image (some other part of data is as second image)) but there are many cycles with different period in my data.
I need to calculate the time of three red, green and purple rectangles for each of the cycles (** the time between each maximum and minimum and total time of cycles **)
Can you give me some ideas on how to do it in R ... is there any special command or package that I can use?
Premise is that the mean value of the data range is used to split the data into categories of peaks and not peaks. Then a running id is generated to group each set of data so an appropriate min or max value can be determined. The half_cycle provides the red and green boxes, while full_cycle provides the purple box for max-to-max and min-to-min. There is likely room for improvement, but it gives a method that can be adjusted as needed.
This sample uses random data since no sample data was provided.
set.seed(7)
wave <- c(seq(20, 50, 10), seq(50, 60, 0.5), seq(50, 20, -10))
df1 <- data.frame(time = seq_len(length(wave) * 5),
data = as.vector(replicate(5, wave + rnorm(length(wave), sd = 5))))
library(dplyr)
df1 %>%
mutate(peak = data > mean(range(df1$data))) %>%
mutate(run = cumsum(peak != lag(peak, default = TRUE))) %>%
group_by(run) %>%
mutate(max = max(data), min = min(data)) %>%
filter((peak == TRUE & data == max) | (peak == FALSE & data == min)) %>%
mutate(max = if_else(data == max, max, NULL), min = if_else(data == min, min , NULL)) %>%
ungroup() %>%
mutate(half_cycle = time - lag(time), full_cycle = time - lag(time, n = 2L))
# A tibble: 11 x 8
time data peak run max min half_cycle full_cycle
<int> <dbl> <lgl> <int> <dbl> <dbl> <int> <int>
1 2 24.0 FALSE 1 NA 24.0 NA NA
2 12 67.1 TRUE 2 67.1 NA 10 NA
3 29 15.1 FALSE 3 NA 15.1 17 27
4 54 68.5 TRUE 4 68.5 NA 25 42
5 59 20.8 FALSE 5 NA 20.8 5 30
6 80 70.6 TRUE 6 70.6 NA 21 26
7 87 18.3 FALSE 7 NA 18.3 7 28
8 108 63.1 TRUE 8 63.1 NA 21 28
9 117 13.8 FALSE 9 NA 13.8 9 30
10 140 64.5 TRUE 10 64.5 NA 23 32
11 145 22.4 FALSE 11 NA 22.4 5 28
I was trying to plot a time series composed of weekly averanges. Here is the plot that I have obtained:
[weekly averages A]
[1]: https://i.stack.imgur.com/XMGMs.png
As you can see the time serie do not cover all the years completely, so, when I have got no data ggplot links two subsequent years. I think I have to group the data in some ways, but I do not understand how. Here is the code:
df4 <- data.frame(df$Date, df$A)
colnames(df4)<- c("date","A")
df4$date <- as.Date(df4$date,"%Y/%m/%d")
df4$week_day <- as.numeric(format(df4$date, format='%w'))
df4$endofweek <- df4$date + (6 - df4$week_day)
week_aveA <- df4 %>%
group_by(endofweek) %>%
summarise_all(list(mean=mean), na.rm=TRUE) %>%
na.omit()
g1 = ggplot() +
geom_step(data=week_aveA, aes(group = 1, x = (endofweek), y = (A_mean)), colour="gray25") +
scale_y_continuous(expand = c(0, 0), limits = c(0, 2500)) +
scale_x_date(breaks="year", labels=date_format("%Y")) +
labs(y = expression(A~ ~index),
x = NULL) +
theme(axis.text.x = element_text(size=10),
axis.title = element_text(size=10))
Here an extraction (the former three years) of the dataset:
endofweek date_mean A_mean week_day_mean
1 20/03/2010 17/03/2010 939,2533437 3
2 27/03/2010 24/03/2010 867,3620121 3
3 03/04/2010 31/03/2010 1426,791222 3
4 10/04/2010 07/04/2010 358,5698314 3
5 17/04/2010 13/04/2010 301,1815352 2
6 24/04/2010 21/04/2010 273,4922895 3,333333333
7 01/05/2010 28/04/2010 128,5989633 3
8 08/05/2010 05/05/2010 447,8858881 3
9 15/05/2010 12/05/2010 387,9828891 3
10 22/05/2010 19/05/2010 138,0770986 3
11 29/05/2010 26/05/2010 370,2147933 3
12 05/06/2010 02/06/2010 139,0451791 3
13 12/06/2010 09/06/2010 217,1286356 3
14 19/06/2010 16/06/2010 72,36972411 3
15 26/06/2010 23/06/2010 282,2911902 3
16 03/07/2010 30/06/2010 324,3215936 3
17 10/07/2010 07/07/2010 210,568691 3
18 17/07/2010 14/07/2010 91,76930829 3
19 24/07/2010 21/07/2010 36,4211218 3,666666667
20 31/07/2010 28/07/2010 37,53981103 3
21 07/08/2010 04/08/2010 91,33282642 3
22 14/08/2010 11/08/2010 28,38587352 3
23 21/08/2010 18/08/2010 58,72836406 3
24 28/08/2010 24/08/2010 102,1050612 2,5
25 04/09/2010 02/09/2010 13,45357513 4,5
26 11/09/2010 08/09/2010 51,24017212 3
27 18/09/2010 15/09/2010 159,7395663 3
28 25/09/2010 21/09/2010 62,71136678 2
29 02/04/2011 31/03/2011 1484,661164 4
30 09/04/2011 06/04/2011 656,1827964 3
31 16/04/2011 13/04/2011 315,3097313 3
32 23/04/2011 20/04/2011 293,2904042 3
33 30/04/2011 26/04/2011 255,7517519 2,4
34 07/05/2011 04/05/2011 360,7035289 3
35 14/05/2011 11/05/2011 342,0902797 3
36 21/05/2011 18/05/2011 386,1380421 3
37 28/05/2011 24/05/2011 418,9624807 2,833333333
38 04/06/2011 01/06/2011 112,7568 3
39 11/06/2011 08/06/2011 85,17855619 3,2
40 18/06/2011 15/06/2011 351,8714638 3
41 25/06/2011 22/06/2011 139,7936898 3
42 02/07/2011 29/06/2011 68,57716191 3,6
43 09/07/2011 06/07/2011 62,31823822 3
44 16/07/2011 13/07/2011 80,7328917 3
45 23/07/2011 20/07/2011 114,9475331 3
46 30/07/2011 27/07/2011 90,13118758 3
47 06/08/2011 03/08/2011 43,29372258 3
48 13/08/2011 10/08/2011 49,39935204 3
49 20/08/2011 16/08/2011 133,746822 2
50 03/09/2011 31/08/2011 76,03928942 3
51 10/09/2011 05/09/2011 27,99834637 1
52 24/03/2012 23/03/2012 366,2625797 5,5
53 31/03/2012 28/03/2012 878,8535513 3
54 07/04/2012 04/04/2012 1029,909052 3
55 14/04/2012 11/04/2012 892,9163416 3
56 21/04/2012 18/04/2012 534,8278693 3
57 28/04/2012 25/04/2012 255,1177585 3
58 05/05/2012 02/05/2012 564,5280546 3
59 12/05/2012 09/05/2012 767,5018168 3
60 19/05/2012 16/05/2012 516,2680148 3
61 26/05/2012 23/05/2012 241,2113073 3
62 02/06/2012 30/05/2012 863,6123397 3
63 09/06/2012 06/06/2012 201,2019288 3
64 16/06/2012 13/06/2012 222,9955486 3
65 23/06/2012 20/06/2012 91,14166632 3
66 30/06/2012 27/06/2012 26,93145693 3
67 07/07/2012 04/07/2012 67,32183278 3
68 14/07/2012 11/07/2012 46,25297513 3
69 21/07/2012 18/07/2012 81,34359825 3,666666667
70 28/07/2012 25/07/2012 49,59130851 3
71 04/08/2012 01/08/2012 44,13438077 3
72 11/08/2012 08/08/2012 30,15773151 3
73 18/08/2012 15/08/2012 57,47256772 3
74 25/08/2012 22/08/2012 31,9109555 3
75 01/09/2012 29/08/2012 52,71058484 3
76 08/09/2012 04/09/2012 24,52495229 2
77 06/04/2013 01/04/2013 1344,388042 1,5
78 13/04/2013 10/04/2013 1304,838687 3
79 20/04/2013 17/04/2013 892,620141 3
80 27/04/2013 24/04/2013 400,1720434 3
81 04/05/2013 01/05/2013 424,8473083 3
82 11/05/2013 08/05/2013 269,2380208 3
83 18/05/2013 15/05/2013 238,9993749 3
84 25/05/2013 22/05/2013 128,4096151 3
85 01/06/2013 29/05/2013 158,5576121 3
86 08/06/2013 05/06/2013 175,2036942 3
87 15/06/2013 12/06/2013 79,20250839 3
88 22/06/2013 19/06/2013 126,9065428 3
89 29/06/2013 26/06/2013 133,7480108 3
90 06/07/2013 03/07/2013 218,0092943 3
91 13/07/2013 10/07/2013 54,08460936 3
92 20/07/2013 17/07/2013 91,54285041 3
93 27/07/2013 24/07/2013 44,64567928 3
94 03/08/2013 31/07/2013 229,5067999 3
95 10/08/2013 07/08/2013 49,70729373 3
96 17/08/2013 14/08/2013 53,38618335 3
97 24/08/2013 21/08/2013 217,2800997 3
98 31/08/2013 28/08/2013 49,43590136 3
99 07/09/2013 04/09/2013 64,88783029 3
100 14/09/2013 11/09/2013 11,04300773 3
So at the end I have one mainly question: how can I eliminated the connection between the years? ... and an aesthetic question: how can I add minor ticks on the x_axis? At least one every 6 months, just to make the plot easy to read.
Thanks in advance for any suggestion!
Edit
This is the code I tried with the suggestion, maybe I mistype some part of it.
library(tidyverse)
library(dplyr)
library(lubridate)
df4 <- data.frame(df$Date, df$A)
colnames(df4)<- c("date","A")
df4$date <- as.Date(df4$date,"%Y/%m/%d")
df4$week_day <- as.numeric(format(df4$date, format='%w'))
df4$endofweek <- df4$date + (6 - df4$week_day)
week_aveA <- df4 %>%
group_by(endofweek) %>%
summarise_all(list(mean=mean), na.rm=TRUE) %>%
na.omit()
week_aveA$endofweek <- as.Date(week_aveA$endofweek,"%d/%m/%Y")
week_aveA$A_mean <- as.numeric(gsub(",", ".", week_aveA$A_mean))
week_aveA$week_day_mean <- as.numeric(gsub(",", ".", week_aveA$week_day_mean))
week_aveA$year <- format(week_aveA$endofweek, "%Y")
library(ggplot2)
library(methods)
library(scales)
mylabel <- function(x) {
ifelse(grepl("-07-01$", x), "", format(x, "%Y"))
}
ggplot() +
geom_step(data=week_aveA, aes(x = endofweek, y = A_mean, group = year), colour="gray25") +
scale_y_continuous(expand = c(0, 0), limits = c(0, 2500)) +
scale_x_date(breaks="6 month", labels = mylabel) +
labs(y = expression(A~ ~index),
x = NULL) +
theme(axis.text.x = element_text(size=10),
axis.title = element_text(size=10))
You have to group by year:
Add a variable with the year to your dataset
Map the year variable on the group aesthetic
For the ticks. Increase the number of the breaks. If you want only ticks but not labels you can use a custom function to get rid of unwanted labels, e.g. my approach below set the breaks to "6 month" but replaces the mid-year labels with an empty string:
week_aveA$endofweek <- as.Date(week_aveA$endofweek,"%d/%m/%Y")
week_aveA$A_mean <- as.numeric(gsub(",", ".", week_aveA$A_mean))
week_aveA$week_day_mean <- as.numeric(gsub(",", ".", week_aveA$week_day_mean))
week_aveA$year <- format(week_aveA$endofweek, "%Y")
library(ggplot2)
mylabel <- function(x) {
ifelse(grepl("-07-01$", x), "", format(x, "%Y"))
}
ggplot() +
geom_step(data=week_aveA, aes(x = endofweek, y = A_mean, group = year), colour="gray25") +
scale_y_continuous(expand = c(0, 0), limits = c(0, 2500)) +
scale_x_date(breaks="6 month", labels = mylabel) +
labs(y = expression(A~ ~index),
x = NULL) +
theme(axis.text.x = element_text(size=10),
axis.title = element_text(size=10))
I want to create new columns that puts TRUE if the number of consecutive wins are two, three etc. So I would like row 3, 6, 7, 8 to be TRUE in a new column called "twoconswins" and row 7, 8 to be true in a new column called "threeconswins" and so on. What is the best way for doing this?
> id date team teamscore opponent opponentscore home win
>9 9 2005-10-05 DET 5 STL 1 1 TRUE
>38 38 2005-10-09 DET 6 CAL 3 1 TRUE
>48 48 2005-10-10 DET 2 VAN 4 1 FALSE
>88 88 2005-10-17 DET 3 SJS 2 1 TRUE
>110 110 2005-10-21 DET 3 ANA 2 1 TRUE
>148 148 2005-10-27 DET 5 CHI 2 1 TRUE
>179 179 2005-11-01 DET 4 CHI 1 1 TRUE
>194 194 2005-11-03 DET 3 EDM 4 1 FALSE
>212 212 2005-11-05 DET 1 PHO 4 1 FALSE
I assumed row 1 should be the header, so that actually rows 2, 5, 6 and 7 should evaluate to TRUE for "twoconswins", and row 6 and 7 for "threeconswins".
You could do:
library(data.table)
df$twoconswins <- (df$win & shift(df$win, 1, NA)) == TRUE
df$threeconswins <- (df$win & shift(df$win, 1, NA) & shift(df$win, 2, NA)) == TRUE
I am thinking this could be more vectorized though, especially if 50 consecutive wins could be possible as well and you'd like to create columns for that as well.
If you like to automatically make the new columns as well, in case it happens sometimes 500 consecutive wins occur, you could do this:
df <- read.table(text =
'id date team teamscore opponent opponentscore home win
9 9 2005-10-05 DET 5 STL 1 1 TRUE
38 38 2005-10-09 DET 6 CAL 3 1 TRUE
48 48 2005-10-10 DET 2 VAN 4 1 FALSE
88 88 2005-10-17 DET 3 SJS 2 1 TRUE
110 110 2005-10-21 DET 3 ANA 2 1 TRUE
148 148 2005-10-27 DET 5 CHI 2 1 TRUE
179 179 2005-11-01 DET 4 CHI 1 1 TRUE
194 194 2005-11-03 DET 3 EDM 4 1 FALSE
212 212 2005-11-05 DET 1 PHO 4 1 FALSE',
header = TRUE)
rles <- data.frame(values = c(rle(df$win)$values),
lengths = c(rle(df$win)$lengths))
maxconwins <- max(rles[rles$values == TRUE,])
for(x in 1: maxconwins){
x <- seq(1,x)
partialstring <- paste("shift(df$win,", x, ",NA)", collapse = " & ")
fullstring <- paste0("df$nr", max(x), "conswins <- (", partialstring, ") == TRUE")
eval(parse(text = fullstring))
}
df[1:maxconwins,9:12][upper.tri(df[1:maxconwins,9:12], diag = TRUE)] <- NA
> df[,8:12]
win nr1conswins nr2conswins nr3conswins nr4conswins
9 TRUE NA NA NA NA
38 TRUE TRUE NA NA NA
48 FALSE TRUE TRUE NA NA
88 TRUE FALSE FALSE FALSE NA
110 TRUE TRUE FALSE FALSE FALSE
148 TRUE TRUE TRUE FALSE FALSE
179 TRUE TRUE TRUE TRUE FALSE
194 FALSE TRUE TRUE TRUE TRUE
212 FALSE FALSE FALSE FALSE FALSE
BTW, I only added the last line because (FALSE & TRUE & TRUE & NA) == TRUE evaluates to FALSE, while you probably like these cells to be NA. I just made sure of this here by setting the upper triagonal of the symmetric submatrix to NA afterwards. For readibility I manually added the column numbers 9 and 12 in here, but you could specify those with a function as well if you'd like.
UPDATE:
When using the Reduce() function as suggested by Frank, you could do this for loop instead of the above:
for(x in 1: maxconwins){
x <- seq(1,x)
eval(parse(text = paste0("df$nr", max(x), "conswins <- (Reduce(`&`, shift(df$win, 1:", max(x), "))) == TRUE")))
}
I am using both geom_hist and histogram in R with the same breakpoints but I get different graphs. I did a quick search, does anyone know what the definition breaks are and why they would be a difference
These produce two different plots.
set.seed(25)
data <- data.frame(Mos=rnorm(500, mean = 25, sd = 8))
data$Mos<-round(data$Mos)
pAge <- ggplot(data, aes(x=Mos))
pAge + geom_histogram(breaks=seq(0, 50, by = 2))
hist(data$Mos,breaks=seq(0, 50, by = 2))
Thanks
To get the same histogram in ggplot2 you specify the breaks inside scale_x_continuous and binwidth inside geom_histogram.
Additionally, hist and histograms in ggplot2 use different defaults to create the intervals:
hist: right-closed (left open) intervals. Default: right = TRUE
stat_bin (ggplot2): left-closed (right open) intervals. Default: right = FALSE
**hist** **ggplot2**
freq1 Freq freq2 Freq
1 (0,2] 0 [0,2) 0
2 (2,4] 2 [2,4) 2
3 (4,6] 2 [4,6) 1
4 (6,8] 1 [6,8) 2
5 (8,10] 6 [8,10) 2
6 (10,12] 9 [10,12) 7
7 (12,14] 24 [12,14) 17
8 (14,16] 27 [14,16) 26
9 (16,18] 39 [16,18) 31
10 (18,20] 48 [18,20) 46
11 (20,22] 52 [20,22) 43
12 (22,24] 38 [22,24) 57
13 (24,26] 44 [24,26) 36
14 (26,28] 46 [26,28) 52
15 (28,30] 39 [28,30) 39
16 (30,32] 31 [30,32) 33
17 (32,34] 30 [32,34) 26
18 (34,36] 24 [34,36) 29
19 (36,38] 18 [36,38) 27
20 (38,40] 9 [38,40) 12
21 (40,42] 5 [40,42) 6
22 (42,44] 4 [42,44) 0
23 (44,46] 1 [44,46) 5
24 (46,48] 1 [46,48) 0
25 (48,50] 0 [48,50) 1
I included the argument right = FALSE so the histogram intervalss are left-closed (right open) as they are in ggplot2. I added the labels in both plots, so it is easier to check the intervals are the same.
ggplot(data, aes(x = Mos))+
geom_histogram(binwidth = 2, colour = "black", fill = "white")+
scale_x_continuous(breaks = seq(0, 50, by = 2))+
stat_bin(binwidth = 2, aes(label=..count..), vjust=-0.5, geom = "text")
hist(data$Mos,breaks=seq(0, 50, by = 2), labels =TRUE, right =FALSE)
To check the frequencies in each bin:
freq <- cut(data$Mos, breaks = seq(0, 50, by = 2), dig.lab = 4, right = FALSE)
as.data.frame(table(frecuencias))
I want to merge two legends in ggplot2. I use the following code:
ggplot(dat_ribbon, aes(x = x)) +
geom_ribbon(aes(ymin = ymin, ymax = ymax,
group = group, fill = "test4 test5"), alpha = 0.2) +
geom_line(aes(y = y, color = "Test2"), data = dat_m) +
scale_colour_manual(values=c("Test2" = "white", "test"="black", "Test3"="red")) +
scale_fill_manual(values = c("test4 test5"= "dodgerblue4")) +
theme(legend.title=element_blank(),
legend.position = c(0.8, 0.85),
legend.background = element_rect(fill="transparent"),
legend.key = element_rect(colour = 'purple', size = 0.5))
The output is shown below. There are two problems:
When I use two or more words in the fill legend, the alignment becomes wrong
I want to merge the two legends into one, such that the fill legend is just part of a block of 4.
Does anyone know how I can achieve this?
Edit: reproducible data:
dat_m <- read.table(text="x quantile y group
1 1 50 0.4967335 0
2 2 50 0.4978249 0
3 3 50 0.5113562 0
4 4 50 0.4977866 0
5 5 50 0.5013287 0
6 6 50 0.4997994 0
7 7 50 0.4961121 0
8 8 50 0.4991302 0
9 9 50 0.4976087 0
10 10 50 0.5011666 0")
dat_ribbon <- read.table(text="
x ymin group ymax
1 1 0.09779713 40 0.8992385
2 2 0.09979283 40 0.8996875
3 3 0.10309222 40 0.9004759
4 4 0.10058433 40 0.8985366
5 5 0.10259125 40 0.9043807
6 6 0.09643109 40 0.9031940
7 7 0.10199870 40 0.9022920
8 8 0.10018253 40 0.8965690
9 9 0.10292754 40 0.9010934
10 10 0.09399359 40 0.9053067
11 1 0.20164694 30 0.7974174
12 2 0.20082056 30 0.7980642
13 3 0.20837821 30 0.8056074
14 4 0.19903399 30 0.7973723
15 5 0.19903322 30 0.8050146
16 6 0.19965049 30 0.8051922
17 7 0.20592719 30 0.8042850
18 8 0.19810139 30 0.7956606
19 9 0.20537392 30 0.8007527
20 10 0.19325158 30 0.8023044
21 1 0.30016463 20 0.6953927
22 2 0.29803646 20 0.6976961
23 3 0.30803808 20 0.7048137
24 4 0.30045448 20 0.6991248
25 5 0.29562249 20 0.7031225
26 6 0.29647060 20 0.7043499
27 7 0.30159103 20 0.6991356
28 8 0.30369025 20 0.6949053
29 9 0.30196483 20 0.6998127
30 10 0.29578036 20 0.7015861
31 1 0.40045725 10 0.5981147
32 2 0.39796299 10 0.5974115
33 3 0.41056038 10 0.6057062
34 4 0.40046287 10 0.5943157
35 5 0.39708008 10 0.6014512
36 6 0.39594129 10 0.6011162
37 7 0.40052411 10 0.5996186
38 8 0.40128517 10 0.5959748
39 9 0.39917658 10 0.6004600
40 10 0.39791453 10 0.5999168")
You are not using ggplot2 according to its philosophy. That makes things difficult.
ggplot(dat_ribbon, aes(x = x)) +
geom_ribbon(aes(ymin = ymin, ymax = ymax, group = group, fill = "test4 test5"),
alpha = 0.2) +
geom_line(aes(y = y, color = "Test2"), data = dat_m) +
geom_blank(data = data.frame(x = rep(5, 4), y = 0.5,
group = c("test4 test5", "Test2", "test", "Test3")),
aes(y = y, color = group, fill = group)) +
scale_color_manual(name = "combined legend",
values=c("test4 test5"= NA, "Test2" = "white",
"test"="black", "Test3"="red")) +
scale_fill_manual(name = "combined legend",
values = c("test4 test5"= "dodgerblue4",
"Test2" = NA, "test"=NA, "Test3"=NA))