cannot estimate contrast

[atnplab]

Hi everybody

I get this error with estimate_contrasts()

estimate_contrasts(fit)
Error: 'summarise_posteriors' is not an exported object from 'namespace:parameters'

How can I get contrasts from stan_lmer object ?

Many thanks

[DominiqueMakowski]

Hey @atnplab

This works at my place:

library(rstanarm)
library(estimate)

data <- iris
data$Petal.Length_factor <- ifelse(data$Petal.Length < 4.2, "A", "B")

model <- stan_glm(Sepal.Width ~ Species * Petal.Length_factor, data = data)

estimate_contrasts(model)
#>            Level1         Level2      Median CI     CI_low     CI_high
#> 2      setosa - A versicolor - A  0.87907014 90  0.7405606  1.02965787
#> 4      setosa - A  virginica - A  0.25415541 90 -1.1625891  1.56523313
#> 1      setosa - A     setosa - B -0.06049493 90 -1.1104129  1.06161341
#> 3      setosa - A versicolor - B  0.51871134 90  0.3942715  0.63901649
#> 5      setosa - A  virginica - B  0.45157749 90  0.3389040  0.55438521
#> 10 versicolor - A  virginica - A -0.62860174 90 -2.0106633  0.72773373
#> 8  versicolor - A     setosa - B -0.94145831 90 -2.0989002  0.11072150
#> 9  versicolor - A versicolor - B -0.36054770 90 -0.5164889 -0.20987725
#> 11 versicolor - A  virginica - B -0.42724056 90 -0.5696512 -0.28055313
#> 13  virginica - A     setosa - B -0.28470296 90 -2.4369123  1.71776140
#> 14  virginica - A versicolor - B  0.27006933 90 -1.0647180  1.67885562
#> 15  virginica - A  virginica - B  0.19970275 90 -1.1692018  1.57588281
#> 6      setosa - B versicolor - B  0.57687244 90 -0.5191995  1.67262806
#> 7      setosa - B  virginica - B  0.51315190 90 -0.6588046  1.52797881
#> 12 versicolor - B  virginica - B -0.06666191 90 -0.1889906  0.05896835
#>         pd ROPE_CI ROPE_low ROPE_high ROPE_Percentage
#> 2  1.00000      89     -0.1       0.1        0.000000
#> 4  0.61500      89     -0.1       0.1        9.688290
#> 1  0.54225      89     -0.1       0.1       14.097164
#> 3  1.00000      89     -0.1       0.1        0.000000
#> 5  1.00000      89     -0.1       0.1        0.000000
#> 10 0.77500      89     -0.1       0.1        7.750632
#> 8  0.91875      89     -0.1       0.1        4.858186
#> 9  1.00000      89     -0.1       0.1        0.000000
#> 11 1.00000      89     -0.1       0.1        0.000000
#> 13 0.60100      89     -0.1       0.1        7.413648
#> 14 0.63050      89     -0.1       0.1       10.193766
#> 15 0.59600      89     -0.1       0.1       10.306094
#> 6  0.80575      89     -0.1       0.1        9.238978
#> 7  0.77275      89     -0.1       0.1       10.643078
#> 12 0.81775      89     -0.1       0.1       69.165965

^{Created on 2019-07-15 by the reprex package (v0.3.0)}

Make sure you have the latest versions of the dependencies (estimate, as well as report, are still unstable due to the work currently focused on their lower dependencies):

install.packages("devtools")

devtools::install_github("easystats/insight")
devtools::install_github("easystats/bayestestR")
devtools::install_github("easystats/parameters", ref="dev")
devtools::install_github("easystats/estimate", ref = "dev")

[atnplab]

I got them in this way, is it correct?

fit <- stan_glmer(score_tot ~ group * RK_tot + (1|id), data = df)
model_parameters(fit)

rope_range <- rope_range(fit)
rope_range

c1 <- emmeans(fit, ~group| RK_tot)
c1
pairs(c1, by = "RK_tot")
c1 <- contrast(c1, interaction = c("pairwise","pairwise"))
c1

c2 <- emmeans(fit, ~RK_tot| group)
c2
pairs(c2, by = "group")
c2 <- contrast(c2, interaction = c("pairwise","pairwise"))
c2

c <- rbind(c1, c2)
c

describe_posterior(c, estimate = "median", dispersion = TRUE, ci = .9, ci_method = "hdi", test = c("p_direction","rope"), rope_range = rope_range)

[DominiqueMakowski]

Nice direct use of the dependencies, yes that should work ☺️

[humanfactors]

I have some code that can also report the contrasts... Let me find that!

[DominiqueMakowski]

With the latest (github) versions of insight, bayestestR and parameters, everything should be fixed ☺️