How to plot distinct regression models using the ggplot2 + ggpmisc or gridExtra packages?

Question

Hello!

I have some difficulty generating graphs with different regression models (quadratic, linear) (FIGURE 1), it is observed that the graphics are loose ("free" scales) when using the grid.arrange(), with the need previously to generate 2 graphs, with their respective regression model, requiring time.

FIGURE 1 - USING grid.arrange

Normally using the ggplot2::ggplot(), i make use of package function ggpmisc::stat_poly_eq(), because, it already organizes the information, and plots, without the need to keep defining the coordinates (x,y). but I cannot list two regression models, at the command stat_poly_eq(formula=). Making a grid of graphics facet_grid() (Fator1 x Fator2 x ... Fatorn), present the same adjustment model for all combinations (FIGURE 2).

FIGURE 2 - Using ggplot()+facet_grid()

Doubts:

1. How to use a fixed cant, for the x-axis, or y, for the grid.arrange? (in this case would be to remove the excess of information, such as axis legends, and numbers on the axis.)

2. How to use two regression adjustment models in the ggplot2 using the ggpmisc::stat_poly_eq()? (in this case would only be to use a ggplot(), avoiding having to generate 2 graphics and combine them with the grid.arrange).

3. How to manually edit the equation to be plotted by ggpmisc::stat_poly_eq()? (in this case I could generate using expression() the distinct models, and add the significance of their Betas).

Database, used:

    require(ggplot2)
    require(ggpmisc)
    require(gridExtra)

base1<-structure(list(FAT1=c(1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,
1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,
1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,1L,
1L,1L,1L,1L,1L,1L),VAR.Y=c(-0.8,-0.75,-0.473,-1.103,
-0.72,-0.667,-0.453,-0.58,-1.327,-0.94,-0.507,-0.68,-0.66,
-0.517,-0.61,-0.893,-1.007,-0.847,-0.767,-0.753,-0.5,
-0.9,-1.24,-1.5,-0.387,-0.4,-0.673,-0.587,-0.79,-0.6,
-0.453,-0.353,-0.413,-0.84,-1.5,-0.763,-0.453,-0.753,
-0.607,-1.1,-0.647,-0.88,-0.513,-0.717,-0.52,-1.093,-1.36,
-0.507),VAR.X=c(0.6193,0.5696,0.5252,0.5643,0.542,0.5694,
0.6386,0.5671,0.5023,0.5626,0.5039,0.5501,0.5966,0.5771,
0.478,0.5855,0.5473,0.5605,0.6068,0.5402,0.4239,0.5775,
0.5254,0.541,0.6267,0.5054,0.5453,0.5699,0.4933,0.5424,
0.5557,0.6236,0.5589,0.5628,0.5364,0.5947,0.5329,0.5283,
0.5062,0.5492,0.4803,0.5593,0.64,0.5602,0.5339,0.5546,
0.5138,0.5451)),class="data.frame",row.names=c(1L,2L,
3L,4L,5L,6L,13L,14L,15L,16L,17L,18L,25L,26L,27L,
28L,29L,30L,37L,38L,39L,40L,41L,42L,49L,50L,51L,52L,
53L,54L,61L,62L,63L,64L,65L,66L,73L,74L,75L,76L,77L,
78L,85L,86L,87L,88L,89L,90L))


base2<-structure(list(FAT1=c(2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,
2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,
2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,2L,
2L,2L,2L,2L,2L,2L),VAR.Y=c(-0.64,-0.8,-0.46,-0.493,
-0.453,-0.387,-0.493,-0.567,-0.347,-0.6,-0.367,-0.97,
-0.62,-0.587,-0.507,-0.66,-0.413,-0.453,-0.6,-0.62,-0.647,
-1.033,-0.507,-0.55,-0.353,-0.4,-0.32,-0.313,-0.347,-0.367,
-0.487,-0.253,-0.413,-0.553,-0.493,-0.44,-0.767,-0.833,
-0.367,-0.713,-0.36,-0.5,-0.44,-0.293,-0.38,-0.427,-0.473,
-0.773),VAR.X=c(0.8934,0.7384,0.282,0.3243,0.2642,0.3908,
0.7625,0.7539,0.4381,0.7273,0.7282,0.4234,0.8397,0.8045,
0.4524,0.7576,0.7217,0.4341,0.855,0.7121,0.3929,0.7137,
0.7924,0.4006,0.3606,0.5003,0.5113,0.428,0.3586,0.4232,
0.6986,0.5425,0.4975,0.5746,0.4854,0.6243,0.9717,0.6435,
0.8064,0.8789,0.4311,0.4133,0.3477,0.2804,0.3823,0.3729,
0.2647,0.4872)),class="data.frame",row.names=c(7L,8L,
9L,10L,11L,12L,19L,20L,21L,22L,23L,24L,31L,32L,33L,
34L,35L,36L,43L,44L,45L,46L,47L,48L,55L,56L,57L,58L,
59L,60L,67L,68L,69L,70L,71L,72L,79L,80L,81L,82L,83L,
84L,91L,92L,93L,94L,95L,96L))

base3<-rbind(base1,base2)

graf1<-ggplot(base1, aes(y=VAR.Y,x=VAR.X))+facet_grid(.~FAT1)+
  geom_point(color="black")+
  geom_smooth(method="lm", se=FALSE, span = .8,color="black")+
  theme_bw()+lims(y=c(-1.8,0))+
  stat_poly_eq(formula = y~I(x)+I(x^2),
               eq.with.lhs = "italic(hat(Y))~`=`~", 
               aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE)

graf2<-ggplot(base2, aes(y=VAR.Y,x=VAR.X))+facet_grid(.~FAT1)+
  geom_point(color="black")+
  geom_smooth(method="lm", se=FALSE, span = .8,color="black")+
  theme_bw()+lims(y=c(-1.8,0))+
  stat_poly_eq(formula = y~I(x),
               eq.with.lhs = "italic(hat(Y))~`=`~", 
               aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE)

graf3<-grid.arrange(graf1,graf2,ncol=2)

ggplot(base3, aes(y=VAR.Y,x=VAR.X))+facet_grid(.~FAT1)+
  geom_point(color="black")+
  geom_smooth(method="lm", se=FALSE, span = .8,color="black")+
  theme_bw()+lims(y=c(-1.8,0))+
  stat_poly_eq(formula = y~I(x)+I(x^2),
               eq.with.lhs = "italic(hat(Y))~`=`~", 
               aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE)

Answer 1

I believe the question is as follows:.

To have two Fits, just call geom_smooth twice, once for each formula. And then add the two equations, adjusting the coordinates of the second equation so that it is not over the first one. The colors of the adjusted lines and their equations are also the same.

And as you can see it is not necessary to use the grid.arrange, facet_grid solves the problem.

ggplot(base3, aes(x = VAR.X, y = VAR.Y)) + 
  geom_point(color = "black")+
  geom_smooth(method = "lm", 
              formula = y ~ x,
              se = FALSE, 
              color = "black") +
  geom_smooth(method = "lm", 
              formula = y ~ poly(x, 2, raw = TRUE),
              se = FALSE, 
              color = "blue") +
  lims(y = c(-1.8, 0)) +
  stat_poly_eq(formula = y ~ x,
               eq.with.lhs = "italic(hat(Y))~`=`~", 
               aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE) +
  stat_poly_eq(formula = y ~ poly(x, 2, raw = TRUE),
               label.y = 0.9, 
               eq.with.lhs = "italic(hat(Y))~`=`~", 
               color = "blue",
               aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE) +
  theme_bw() +
  facet_grid(. ~ FAT1)

Answer 2

There is no option for each Facet to have a regression type, you have to make two separate graphs. To avoid repetition, you can write a function.

To join the graphics, I used the package ggpubr, that has some amenities in relation to the gridExtra. Since I don’t have the package ggpmisc installed, I’m using the function ggpubr::stat_regline_equation to plot the equations.

library(ggplot2)
library(ggpubr)

plot.reg <- function(df, var, formula) {
  form <- as.formula(paste("y ~", formula))
  ggplot(subset(df, FAT1 == var) , aes(VAR.X, VAR.Y)) +
    facet_grid(.~FAT1) +
    geom_point(color = "black") +
    geom_smooth(method = "lm", formula = form, se = FALSE, span = .8, color = "black") +
    #stat_poly_eq(formula = form,
    #             eq.with.lhs = "italic(hat(Y))~`=`~",
    #             aes(label = paste(..eq.label.., ..rr.label..,  sep =  "*plain(\",\")~")), parse = TRUE) +
    stat_regline_equation(aes(label =  paste(..eq.label.., ..rr.label.., sep = "*plain(\",\")~~")),
                          formula = form) +
    lims(y = c(min(df[["VAR.Y"]]), max(df[["VAR.Y"]])+.2),
         x = c(min(df[["VAR.X"]]), max(df[["VAR.X"]]))) +
    labs(x = NULL, y = NULL) +
    theme_bw()
}

p1 <- plot.reg(base3, 1, "I(x)+I(x^2)")
p2 <- plot.reg(base3, 2, "I(x)")

annotate_figure(ggarrange(p1, p2+rremove("y.text")+rremove("y.ticks"),
                          nrow = 1, widths = c(1.1, 1)),
                left = names(base3)[2],
                bottom = names(base3)[3])

The same can be obtained with gridExtra::grid.arrange, It’s just a matter of personal preference.