Display regression slopes for multiple subsets in ggplot2 (facet_grid)

Question

To my fellow programmers,

I have been searching all over the web for an answer to this question and I am completely stumped.

Quite simply, I am trying to display a slope (y=mx+b) and an R-squared value on my ggplot2 figure (using RStudio).

In my experiment, I measure the response of a bacteria to different media compositions (food sources). Therefore, in one figure, I have many panels (or subsets) and each have a different R^2 and slope. In this example, I have 6 different subsets, all of which have the exact same axes.

To do so, I have created a "MasterTable"

MasterTable <- inner_join(LongRFU, LongOD, by= c("Time.h", "Well", "Conc.nM", "Assay"))

with all of my raw values in a long table format (subsets are called Assay).

Then I created a second table which only displays the R-squared values of the 6 subsets:

Determine the correlation of each subset

Correlation <- ddply(MasterTable, .(Assay), summarise, cor = round(cor(maxRFU, Conc.nM), 3))
Correlation$Rsquared <- (Correlation$cor)^2
Correlation$Rsquared <- round(Correlation$Rsquared,3)

With this command, I have managed to display the R^2 in all of my subsets of my figure (ggplot command bellow).

Calculating the regression's slope and intercept:

slope <- dlply(MasterTable, .(Assay), lm, formula = (maxRFU ~ Conc.nM))
m <- lm(MasterTable$Conc.nM ~ MasterTable$maxRFU)
a <- signif(coef(m)[1], digits = 2)
b <- signif(coef(m)[2], digits = 2)
textlab <- paste("y = ",b,"x + ",a, sep="")
print(textlab[1])

Creating the ggplot

ggplot(data=MasterTable, aes(x=Conc.nM, y=maxRFU)) + 
    geom_point(shape = 21, size = 2, colour = "black", fill = "#606060") + 
    geom_smooth(method = "lm", colour = "#001170", se=TRUE) + 
    geom_text(data=Correlation, 
            aes(label=paste("R^2=", " ", Rsquared, sep="")), x=200, y=550) + 
    annotate("text", x = 200, y = 500, label = textlab, color="black", 
              size = 5, parse=FALSE) + 
    geom_errorbar(aes(ymin=maxRFU-sdRFU, ymax=maxRFU+sdRFU), width=.05, 
                  colour="#4b4b4b", linetype = "solid", size = 0.1) +
    theme(panel.background = element_rect(fill="white", 
          linetype = "solid", colour = "black"), 
          legend.key = element_rect(fill = "white"), 
          panel.grid.minor = element_blank(), panel.grid.major = element_blank()) + 
    facet_grid(. ~ Assay) + 
    labs(title="Calibration curves", y="Max RFU", 
         x="As(III) concentration (nM)")

StackOverflow is not allowing me to post an image yet... So I have uploaded it to my box account:

Calibration curves

Now you see my problem?

All of the subsets display the exact same slope. I cannot seem to find a fix for this and I would really appreciate your help in the matter.

I also have a bonus question. This is a silly one, but would be awesome if someone knows how to fix it. I would really like to display the R^2 value in subscript like it is shown in this link: Subscript in ggplot

Oh and final bonus question: I can't seem to place a "cap" on the error bars...

Thanks again for all of the help in the matter,

Marty

http://stackoverflow.com/questions/19699858/ggplot-adding-regression-line-equation-and-r2-with-facet?answertab=votes#tab-top might help — user20650, Jan 27 '15 at 03:08
I can't be certain without seeing samples of each data frame you used to make the plot (and you should generally post such samples in the future, so that we can run your code and reproduce the problem you're trying to solve), but it looks like you're using `annotate` to add the regression equation text to each plot. Instead, use `geom_text`. As long as the `textlab` data frame has a column for `Assay`, `geom_text` will put the regression equation in each panel that matches each value of `Assay` in `textlab`. — eipi10, Jan 27 '15 at 03:12
For completeness, I should have mentioned in my previous comment that the name of the column containing the text labels needs to go inside `aes` in `geom_text` (just as in your other call to `geom_text`). — eipi10, Jan 27 '15 at 03:25
Thank you eipi! You figured it out. How do I make your post the answer? — Marty999, Jan 27 '15 at 05:36

Marty999 · Accepted Answer · 2015-01-28T03:20:32.267

eipi has figured out the answer. I made a few modifications to my code and it is working now.

Intercept <- ddply(MasterTable, .(Assay),function(x) coefficients(lm(maxRFU~Conc.nM,x)))
names (Intercept) <- c("Assay", "Intercept", "Slope")
Intercept$Intercept <- round(Intercept$Intercept,0)
Intercept$Slope <- round(Intercept$Slope,3)

The ddply line added a table that looked like this:

                Assay   Intercept    Slope
1               B-GMM    601.2766  0.3758356
2       B-GMM + As(V)   1271.3436 -0.5405553
3 B-GMM + As(V) + PO4    699.9420  0.8970737
4         B-GMM + PO4    720.5684  1.0486098
5                 GMM    749.9787  1.9294352
6         GMM + As(V)    768.1253  1.4962517

Then, in the ggplot, I added the following line:

geom_text(data=Intercept, aes(label=paste("y = ",Slope,"x + ",Intercept, sep="")), x=200, y=500)

As simple as that.

Thanks again!

Glad you were able to get the plot working the way you wanted. You should go ahead and accept your answer (although you might have to wait a day or two for the site to let you accept an answer to your own question). — eipi10, Jan 27 '15 at 19:51

Display regression slopes for multiple subsets in ggplot2 (facet_grid)

Determine the correlation of each subset

Calculating the regression's slope and intercept:

Creating the ggplot

1 Answers1

Linked