Plotting different models

profgaelle

Hello!

I'm wondering how I could reproduce the two-panels figure comparing a model with interaction and a model without interaction on the Bayes Factor page:
https://richarddmorey.github.io/BayesFactor/#glm
Anyone knows the r code for plotting this?

Thank you in advance!

richarddmorey

The basic steps are:

1. Make a scatterplot
2. Fit the model
3. Extract the model parameters from the object (whatever it is)
4. Use abline() to plot the lines

You can do this with either lm() estimates or BayesFactor::posterior() estimates (or whatever). I'll demonstrate below with lm() because it is slightly easier.

# scatter plot
plot(len ~ dose, data = ToothGrowth, log = "x", 
     pch = as.integer(ToothGrowth$supp)+20,
     bg = as.integer(ToothGrowth$supp), 
     las = 1)

# Use log10 because I plotted the log of x above
cf = coef(lm(len ~ log10(dose)*supp, data = ToothGrowth))
# coef extracts the coefficients
abline(a = cf["(Intercept)"], b = cf["log10(dose)"])
abline(a = cf["(Intercept)"] + cf["suppVC"], 
       b = cf["log10(dose)"] + cf["log10(dose):suppVC"], col= "red")


## Now for the no-interaction model
# scatter plot
plot(len ~ dose, data = ToothGrowth, log = "x", 
     pch = as.integer(ToothGrowth$supp)+20,
     bg = as.integer(ToothGrowth$supp), 
     las = 1)

# Use log10 because I plotted the log of x above
cf = coef(lm(len ~ log10(dose) + supp, data = ToothGrowth))
# coef extracts the coefficients
abline(a = cf["(Intercept)"], b = cf["log10(dose)"])
abline(a = cf["(Intercept)"] + cf["suppVC"], 
       b = cf["log10(dose)"], col= "red")

MSB

For many supported models, you can also the tidyverse approach, with the ggeffects package.