BF_inclusion sensitive to order of terms in an interaction

[mattansb]

In the example below, shape:color and color:shape should be one term (same for color:shape:ID and shape:color:ID):

library(bayestestR)
library(lme4)


data("puzzles", package = "BayesFactor")
puzzles <- as.data.frame(lapply(puzzles, function(x) rep(x,5)))
puzzles$RT <- puzzles$RT + rnorm(240)

m1 <- lmer(RT ~ shape + shape:color + (shape + shape:color|ID), puzzles)
m2 <- lmer(RT ~ color + shape:color + (color + shape:color|ID), puzzles)
m3 <- lmer(RT ~ color:shape + (color:shape|ID), puzzles)

x <- bayesfactor_models(m1,m2,m3)
bayesfactor_inclusion(x)
#> # Inclusion Bayes Factors (Model Averaged)
#> 
#>                Pr(prior) Pr(posterior) Inclusion BF
#> shape               0.33           0.5            2
#> shape:color         0.33           0.5            2
#> 1:ID                1.00           1.0          NaN
#> shape:ID            0.33           0.5            2
#> shape:color:ID      0.33           0.5            2
#> color               0.33           0.5            2
#> color:shape         0.67           0.5          0.5
#> color:ID            0.33           0.5            2
#> color:shape:ID      0.67           0.5          0.5
#> 
#> * Compared among: all models
#> *    Priors odds: uniform-equal

^{Created on 2019-09-10 by the reprex package (v0.3.0)}

[mattansb]

Fixed

library(bayestestR)
library(lme4)

data("puzzles", package = "BayesFactor")
puzzles <- as.data.frame(lapply(puzzles, function(x) rep(x,5)))
puzzles$RT <- puzzles$RT + rnorm(240)

m1 <- lmer(RT ~ shape + shape:color + (shape + shape:color|ID), puzzles)
m2 <- lmer(RT ~ color + shape:color + (color + shape:color|ID), puzzles)
m3 <- lmer(RT ~ color:shape + (color:shape|ID), puzzles)

x <- bayesfactor_models(m1,m2,m3)
bayesfactor_inclusion(x)
#> # Inclusion Bayes Factors (Model Averaged)
#> 
#>                Pr(prior) Pr(posterior) Inclusion BF
#> shape               0.33           0.5            2
#> color:shape         1.00           1.0          NaN
#> 1:ID                1.00           1.0          NaN
#> shape:ID            0.33           0.5            2
#> color:shape:ID      1.00           1.0          NaN
#> color               0.33           0.5            2
#> color:ID            0.33           0.5            2
#> 
#> * Compared among: all models
#> *    Priors odds: uniform-equal

^{Created on 2019-09-10 by the reprex package (v0.3.0)}