rggplot2emmeans

How to change the order of terms on the x axis using emmip or emmip_ggplot?

I need some help. I am trying to change the x axis terms while using the emmip function. Both of the items in question have been coded as factors (Ploidy and P). x is my data.

x$Ploidy <- as.factor(x$Ploidy)
x$P <- as.factor(x$P)
summary(x)

Output of summary:

Test tube number  Sample ID           Lineage                 Ploidy      P         absorb 1      
 Min.   :  1.00   Length:70          Length:70          Diploid   :22   High:34   Min.   :0.04300  
 1st Qu.: 46.25   Class :character   Class :character   Triploid  :21   Low :36   1st Qu.:0.05425  
 Median : 80.50   Mode  :character   Mode  :character   Tetraploid:27             Median :0.05700  
 Mean   : 79.04                                                                   Mean   :0.06189  
 3rd Qu.:109.75                                                                   3rd Qu.:0.06600  
 Max.   :149.00                                                                   Max.   :0.09800  
    absorb 2         absorb avg           +/-          absorb x 1.73        dilution          mg P          
 Min.   :0.04600   Min.   :0.04450   Min.   :0.01270   Min.   :0.07698   Min.   :0.007   Min.   :8.893e-05  
 1st Qu.:0.05500   1st Qu.:0.05613   1st Qu.:0.05577   1st Qu.:0.09710   1st Qu.:0.007   1st Qu.:3.904e-04  
 Median :0.06000   Median :0.05900   Median :0.06642   Median :0.10207   Median :0.007   Median :4.649e-04  
 Mean   :0.06147   Mean   :0.06168   Mean   :0.07634   Mean   :0.10670   Mean   :0.007   Mean   :5.344e-04  
 3rd Qu.:0.06600   3rd Qu.:0.06650   3rd Qu.:0.09420   3rd Qu.:0.11504   3rd Qu.:0.007   3rd Qu.:6.594e-04  
 Max.   :0.08900   Max.   :0.08950   Max.   :0.17940   Max.   :0.15484   Max.   :0.007   Max.   :1.256e-03  
   mg sample         USE mg sample snails corrected   P_content       
 Min.   :0.0000040   Min.   :0.0040                 Min.   :0.000186  
 1st Qu.:0.0002365   1st Qu.:0.2365                 1st Qu.:0.010775  
 Median :0.0004020   Median :0.4020                 Median :0.019333  
 Mean   :0.0004277   Mean   :0.4277                 Mean   :0.024214  
 3rd Qu.:0.0006230   3rd Qu.:0.6230                 3rd Qu.:0.030923  
 Max.   :0.0015900   Max.   :1.5900                 Max.   :0.125460

I ran a generalized linear mixed model and visualized the data using emmip. gammaP is my model.

install.packages("emmeans")
library(emmeans)
emmip(gammaP, P ~ Ploidy, xlab = "Ploidy Level")

I receive the plot: emmip image

I need to switch the order of Triploid and Tetraploid in the graph. I attempted to do this with this code:

x$Ploidy <- factor(x$Ploidy, levels=c("Diploid", "Triploid", "Tetraploid"))
emmip(gammaP, P ~ Ploidy, xlab = "Ploidy Level")

As you can see the axis labels changed but not the actual points on the plot.

bad emmip plot

Why is only the axis changing? Should I have not recoded my data?

I attempted this again with emmip_ggplot. As emmip_ggplot uses the same syntax as ggplot, but the axis terms nor the points change.

I recoded my data to a character and then made a plot to use emmip_ggplot:

x$Ploidy <- as.character(x$Ploidy)
emmip(gammaP, P ~ Ploidy)
badplot <- emmip(gammaP, P ~ Ploidy, plotit = FALSE)
emmip_ggplot(badplot, P ~ Ploidy, aes(x = factor(Ploidy, level = c('Diploid', 'Triploid', 'Tetraploid'))))

The x axis did not change again and I am concerned that I will only change the axis terms again... more bad plots

Thank you all!

Please let me know if you would like to see a snap of my data or the model!

Solution

  • The order of the categories on the x axis is in general determined by the order of the levels of the categories of a factor. As you have not set the order when converting to a factor you get the default order, i.e. the categories are ordered alphabetically.

    Hence, to fix your issue set your desired order when converting to a factor.

    Adapting the default example from ?emmeans::emmip:

    library(emmeans)
library(ggplot2)

dat <- auto.noise

# Set the order of the levels 
dat$size <- factor(dat$size, c("L", "S", "M"))
noise.lm <- lm(noise ~ size * type, data = dat)

emmip(noise.lm, type ~ size)

    Or as a second option you can set the order of the categories on the x axis using the limits= argument of scale_x_discrete:

    
# Set the order using scale_x_discrete
emmip(noise.lm, type ~ size) +
  scale_x_discrete(limits = c("M", "S", "L"))