Scatter plot (two Y variables and one X variable) with regression line and equation

[Lidwina Bertrand]

Dear alls,
I need help to make a scatter plot with ggplot2.
My data is like:

S	P	SP
0,05146604	20,0546	20,11
0,03857084	20,0546	20,09
0,03121054	12,2288	12,26
0,30647028	12,5126	12,82
0,28482765	14,7220	15,01
0,50051247	45,0246	45,53
0,19338897	59,2978	59,49
1,2906287	86,3796	87,67
0,43907263	79,2731	79,71
0,53855983	77,6721	78,21
0,17065523	39,8464	40,02
0,22226842	10,8266	11,05
0,28976245	14,1776	14,47
0,82201466	13,1247	13,95
0,26347576	21,1123	21,38
0,59649658	27,6495	28,25
0,22606365	12,2622	12,49
0,32395152	24,2651	24,59
0,08511838	24,2651	24,35
0,02545399	15,4228	15,45
0,36901547	51,4263	51,80
0,47847889	37,6949	38,17
3,01828548	70,2802	73,30
0,27841585	58,9307	59,21
0,24509299	31,4420	31,69
0,51293635	6,3886	6,90
0,43961007	20,1171	20,56
0,3532001	10,2026	10,56
1,80320357	40,0416	41,84
0,08934678	15,9644	16,05
0,23344384	119,5648	119,80
0,88881239	86,8860	87,77
0,49942548	115,9956	116,49
0,80560236	38,8773	39,68
1,0598936	47,3107	48,37
6,4852884	166,4927	172,98
0,20561425	81,9291	82,13
1,64810016	197,9828	199,63
1,27276508	17,2545	18,53

I need a scatter plot with S (Y1) vs SP (X) (with a dot color) and P (Y2) vs SP (X) (with an other dot color) in the same plot, a regression line for each set of data and finally the equation with R2.

Could you help me?

thank you

Lidwina

[Dennis Murphy]

Hi:

The general process is this:

1. Fit both of your models outside of ggplot2 and extract the coefficients from each.

2. Create a data frame that contains the x and y positions where you want to place the fitted equations along with a third column that contains a text string representation of the fitted line using plotmath code. You can either write a function to generate this string or produce it manually - for only two lines, the manual approach probably takes less time. There should be one line in this data frame for each model.

3. Melt the data such that your two explanatory variables are stacked. This makes it easier to plot lines with different colors or line types, for example.

4. Create a ggplot using the result of the melt operation in step 3 as the input data. Use geom_smooth(method = "lm")  to generate the two lines and geom_text() to insert the fitted line equations. You'll need to use parse = TRUE as an argument to geom_text() in order to convert the text string to a mathematical equation.

ggplot2 can generate the fitted lines internally, but it does not "remember" the model coefficients, so you can't use ggplot2 directly to get at the fitted coefficients. There is a way to get them after the fact, but it is not any less complicated than fitting the models beforehand and generating a data frame to create and position the fitted lines by hand.

Since your data are unreadable as is (it is never a good idea to copy/paste data from the R console or a spreadsheet into a message to a text-based list), you'll have to figure out several of these steps on your own, but here are a few lines of code to get you started, taking the name of your data frame as DF:

# Get the model coefficients

coef1 <- coef(lm(S ~ SP, data = DF))

coef2 <- coef(lm(P ~ SP, data = DF))

# Generate the data frame for the equations on your own

# See ?plotmath - you'll need at least hat[y], ~, == and +

# The examples of ?plotmath are helpful - there are also

# examples of how to do this with ggplot2 in the list archives.

# How you want to input the coefficients is up to you.

# For the first call to ggplot() below, assume this data frame has columns

# x, y and eqn for the x-position, y-position and equation as text string,

# respectively.

# Melt the data

library(reshape2)

DFm <- melt(DF, id = "SP")    # generates two new vars named variable and value

# Do the plot - puts both lines in the same panel

library(ggplot2)

ggplot(DFm, aes(x = SP, y = value, colour = variable)) +

    geom_point() +

    geom_smooth(method = "lm", se = FALSE, size = 1) +

    geom_text(data = coefDF, aes(x = x, y = y, label = eqn), parse = TRUE) +

    scale_colour_manual(values = c("darkorange", "blue"))

This will give an inane plot because the range of P is much wider than the range of S, so you're better off faceting the two plots into separate panels (no, ggplot2 does not produce multiple y-scales...by design). In this case, you need to add another column to the data frame containing the fitted equation: variable = factor(c("S", "P")),  which represents the y-variable in each row. Then,

# separate plot for each line

ggplot(DFm, aes(x = SP, y = value, colour = variable)) +

    geom_point() +

    geom_smooth(method = "lm", se = FALSE, size = 1) +

    geom_text(data = coefDF, aes(x = x, y = y, label = eqn), parse = TRUE) +

    scale_colour_manual(values = c("darkorange", "blue"))  +

    facet_wrap(~ variable, ncol = 1, scales = "free_y")

Getting the data frame for the fitted equations is the hard part: you'll have to decide where to position each equation and you have to remember that ggplot2 expects correct plotmath code encased in a text string. The argument parse = TRUE parses and evaluates the string before plotting the result.

Dennis

--
--
You received this message because you are subscribed to the ggplot2 mailing list.
Please provide a reproducible example: https://github.com/hadley/devtools/wiki/Reproducibility
 
To post: email ggp...@googlegroups.com
To unsubscribe: email ggplot2+u...@googlegroups.com
More options: http://groups.google.com/group/ggplot2

---
You received this message because you are subscribed to the Google Groups "ggplot2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ggplot2+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Vivek Patil]

This discussion on stack overflow http://stackoverflow.com/questions/7549694/ggplot2-adding-regression-line-equation-and-r2-on-graph can help. I used that and faceting for a blog post at http://patilv.com/Interactive-Anscombe-Viz/