R - ggplot multiple regression lines for different columns in same chart

Question

With data like below,

text = "
R_1700,R_350,R_2950,S_1700,S_350,S_2950
,,-98.2,,,14.15
-80,-82.3,-99,,-0.7,12.4
-77.55,-80.6,-97,,,14.5
-75.55,-80.85,-96.35,,,14.4
-80.8,-81.6,-94.3,,9.95,6
-80.8,-81.8,,,4.9,
-80.8,-81.85,,,8.2,
-73.8,-77.6,-98,,6.35,
-72.8,-76.7,-96.8,3.7,4.6,
-72.65,-81.7,-94.05,2.25,,
-72.95,-80.4,-94.6,1.7,,
-72.7,-81.7,-94.35,1.6,,
-76.05,-84.25,-95.65,3.65,,
-75.5,-84.65,-95.2,1.95,,
-74.65,-83.8,-94.6,2.6,,
-74.2,-83.95,-100.65,3.25,,
-66.8,-75.65,-97.25,,6.45,
-73.7,-77.7,-97.05,,6.8,
-97.8,-100.8,-116.9,,-5.3,
,-99.7,,,-1,
,-100.2,,,-1.3,
-93.3,-94.75,-103.7,,-4.25,
-94.6,-96.55,-105,,-6.7,
-96.4,-98.45,-110.1,,-6.9,
-96.4,-101.1,-110.7,,-7.65,
-94.95,-102,,,-7.2,
-94,-102.15,,,-9.35,
-91.8,-97,-110.3,,-5.3,
"
df1 = read.table(textConnection(text), sep=",", header = T)

Need to plot regression lines for columns as below with X-axis holding R_... values and Y-axis holding S_... values

S_1700 vs. R_1700
S_350 vs. R_350
S_2950 vs. R_2950

For a single group of variables, I could have done something like below.

ggplot(df1, aes(x=R_1700, y=S_1700)) +
  geom_point() + 
  geom_smooth(method=lm, se=FALSE, fullrange=TRUE)

Need help to get all the three lines in a single plot as in the example below. The 3 different groups would be 1700, 350 and 2950.

UseR10085 · Accepted Answer · 2020-07-15 07:14:48Z

2

tidyverse solution

library(tidyverse)

df1 %>% 
  pivot_longer(everything()) %>% #wide to long data format
  separate(name, c("key","number"), sep = "_") %>% #Separate elements like R_1700 into 2 columns 
  group_by(number, key) %>% #Group the vaules according to number, key
  mutate(row = row_number()) %>% #For creating unique IDs 
  pivot_wider(names_from = key, values_from = value) %>% #Make separate columns for R and S
  ggplot(aes(x=R, y=S, color = number, shape = number)) +
  geom_point() + 
  geom_smooth(method=lm, se=FALSE, fullrange=TRUE)

edited Jul 15, 2020 at 7:14

answered Jul 15, 2020 at 6:56

UseR10085

8,3164 gold badges35 silver badges72 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

user3206440 Over a year ago

great .. would you also please add comments to the steps being done? Also in separate you have used sep = "_" - what if the instead of R_1700 its a string pattern like ABC_CA_BDEF_for_KEY_1700 and instead of 'S_1700' is a string pattern like ABC_CA_XYZX_for_KEY_1700. Note that the difference between the two is between the third and fourth underscores

UseR10085 Over a year ago

See the updated answer. For details about separate you can visit this.

UseR10085 Over a year ago

Also visit this

mt1022 · Accepted Answer · 2020-07-15 07:07:42Z

2

If you could reorganize your data in a format like below:

# with data.table package
library(data.table)
setDT(df1)
df2 <- melt(df1, measure.vars = patterns('R_', 'S_'))
df2[, variable := factor(variable, levels = 1:3,
    labels = tstrsplit(grep('R_', names(df1), value = TRUE), '_')[[2]])]
# > df2
#     variable  value1 value2
# 1:     1700      NA     NA
# 2:     1700  -80.00     NA
# 3:     1700  -77.55     NA
# 4:     1700  -75.55     NA
# 5:     1700  -80.80     NA
# 6:     1700  -80.80     NA
# 7:     1700  -80.80     NA
# 8:     1700  -73.80     NA
# 9:     1700  -72.80   3.70


# without data.table
tmp <- split.default(df1, f = sapply(strsplit(names(df1), '_'), `[`, 2))
tmp <- lapply(tmp, function(dtf){
    names(dtf) <- c('value1', 'value2')
    return(dtf)
})
df2 <- do.call(rbind, tmp)
df2$variable <- rep(names(tmp), each = nrow(df1))

you can visualize the data as desired easily:

ggplot(df2, aes(x = value1, y = value2, color = variable)) +
    geom_point() + 
    geom_smooth(method=lm, se=FALSE, fullrange=TRUE) +
    labs(x = 'R', y = 'S')

edited Jul 15, 2020 at 7:07

answered Jul 15, 2020 at 6:34

mt1022

17.4k6 gold badges53 silver badges77 bronze badges

2 Comments

user3206440 Over a year ago

possible to get an answer without data.table ?

mt1022 Over a year ago

@user3206440, of course. see the edited answer. It does matter how you reformat the data, as long as the reorganized data are in long format that are suitable for ggplot. For the current data with just several columns, you can do it manually with column subsetting -> rename->rbind, and finally, adding a column of the grouping variable. But an automatic way is more general and works for dataset with more columns.

Collectives™ on Stack Overflow

R - ggplot multiple regression lines for different columns in same chart

2 Answers 2

3 Comments

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related