get_response(<brms>) fails with aterms

[Chinca-lang]

Hi,
I'm trying to use describe_posterior with a model fitted with brm but I get this error:

describe_posterior(modelSex2, ci=0.89)
Error in [.data.frame(model_data, , rn, drop = FALSE) :
undefined columns selected

My model is:

modelSex2 <- brm(formula= ALI | trunc(lb = 0, ub = 10)~Sex*as.factor(date) + (1|ID),
chains = 10, iter = 5000, warmup = 1000,
refresh = 0,
data = data_as_cal, family = poisson)

I suspect that trunc function is causing some trouble as I've fitted the same model without trunc and describe_posterior is working well.
Do you have any suggestions?

[Chinca-lang]

just a quick update
If I set the rope_range it works:

describe_posterior(modelSex2, ci=0.89, rope_range = rope_range)

where rope_range is:

rope_value <- 0.1*sd(data_as_cal$ALI)
rope_range <- c(-rope_value, rope_value)

I don't know why it was difficult to calculate automatically the default rope range. But this fixed the issue.

[mattansb]

This is actually an issue in insight::get_response().

library(brms)
library(bayestestR)

fit4 <- brm(count | trunc(ub = 104) ~ zBase * Trt,
            data = epilepsy, family = poisson(),
            backend = "cmdstanr", cores = 4)

insight::get_response(fit4) # source = "environment"
#> Error in `[.data.frame`(model_data, , rn, drop = FALSE) : 
#>   undefined columns selected

insight::get_response(fit4, source = "mf")
#> Error in `[.data.frame`(model_data, , rn, drop = FALSE) : 
#>   undefined columns selected

[mattansb]

The issue is a general issue with brms's aterms.
Another example: easystats/performance#596

[mattansb]

When the aterm has a constant (not a data column), you get the error:

library(brms)

fit4 <- brm(am | trials(1) ~ hp,
            data = mtcars, family = binomial(),
            backend = "cmdstanr", cores = 4)

insight::get_response(fit4) # source = "environment"
#> Error in `[.data.frame`(model_data, , rn, drop = FALSE) : 
#>   undefined columns selected

insight::get_response(fit4, source = "mf")
#> Error in `[.data.frame`(model_data, , rn, drop = FALSE) : 
#>   undefined columns selected

This also means that your fix at #781 is wrong for this bug:

1. If the aterm uses a variable from the data, that should be returned (regardless of the aterm type).
2. If is it a constant, don't return it.

Originally posted by @mattansb in easystats/performance#596 (comment)

[SethMusker]

I'm still getting this error (along with other related[?] ones) with a multivariate brms model with missing data imputation.

Example based on the marginaleffects brms vignette (https://marginaleffects.com/vignettes/brms.html)

library(marginaleffects)
library(brms)
dat <- read.csv("https://vincentarelbundock.github.io/Rdatasets/csv/carData/TitanicSurvival.csv")
dat$survived <- ifelse(dat$survived == "yes", 1, 0)
dat$woman <- ifelse(dat$sex == "female", 1, 0) |> as.factor()
sum(is.na(dat$age))  # <-- 263 missing values

mod_miss <- brm(bf(survived ~ woman*mi(age) + passengerClass, family=bernoulli(link = "logit")) +
              bf(age | mi() ~ passengerClass + woman) +
              set_rescor(FALSE),
           data = dat,
           backend = "cmdstanr",
           cores=4)

#--------------------------------------
## Example 1
plot_slopes(mod_miss, 
   variables = "woman", 
   condition = "age")
# Error in `[.data.frame`(model_data, , rn, drop = FALSE) : 
#   undefined columns selected

#--------------------------------------
## Example 2
slopes(mod_miss,
    newdata = datagrid(
    woman = 1,
    passengerClass = "1st"))
# Error: The following variables can neither be found in 'data' nor in 'data2':
#   'age'
# In addition: Warning message:
#   These variables were not found: survived, age, .  Try specifying the `newdata` argument explicitly and make sure
# the missing variable is included.

#--------------------------------------
## Example 3
slopes(mod_miss,
    variables = "age",
    newdata = datagrid(woman = 0:1)) |>
    posterior_draws()
# Error: There is no valid predictor variable. Please change the `variables` argument or supply a new data frame to the
# `newdata` argument.
# In addition: Warning message:
#   These variables were not found: miage.  Try specifying the `newdata` argument explicitly and make sure the
# missing variable is included. 

And the tracebacks.

#--------------------------------------
## Example 1
8: stop("undefined columns selected")
7: `[.data.frame`(model_data, , rn, drop = FALSE)
6: model_data[, rn, drop = FALSE]
5: get_response.default(model)
4: insight::get_response(model)
3: sanitize_condition(model, condition, variables, modeldata = modeldata)
2: plot_comparisons(model, variables = variables, condition = condition, 
       by = by, newdata = newdata, type = type, vcov = vcov, conf_level = conf_level, 
       wts = wts, draw = draw, rug = rug, gray = gray, comparison = slope, 
       ...)
1: plot_slopes(mod_miss, variables = "woman", condition = "age")

#--------------------------------------
## Example 2
7: stop(format_message(string = string, ..., line_length = line_length, 
       indent = indent), call. = call)
6: format_alert(..., type = "error")
5: insight::format_error(pred_both[["error"]][["message"]])
4: get_contrasts(structure(list(formula = structure(list(forms = list(
       survived = structure(list(formula = survived ~ woman * mi(age) + 
           passengerClass, pforms = list(), pfix = list(), family = structure(list(
           family = "bernoulli", link = "logit", linkfun = function (mu) 
           link(mu, link = slink), linkinv = function (eta) 
           inv_link(eta, link = slink), dpars = "mu", type = "int", 
           ybounds = c(0, 1), closed = c(TRUE, TRUE), ad = c("weights", 
           "subset", "index"), specials = c("binary", "sbi_logit"
           )), class = c("brmsfamily", "family")), resp = "survived", 
           mecor = TRUE), class = c("brmsformula", "bform")), age = structure(list(
           formula = age | mi() ~ passengerClass + woman, pforms = list(), 
           pfix = list(), resp = "age", family = structure(list(
               family = "gaussian", link = "identity", linkfun = function (mu) 
               link(mu, link = slink), linkinv = function (eta) 
               inv_link(eta, link = slink), dpars = c("mu", "sigma"
               ), type = "real", ybounds = c(-Inf, Inf), closed = c(NA, 
               NA), ad = c("weights", "subset", "se", "cens", "trunc", 
               "mi", "index"), normalized = c("_time_hom", "_time_het", 
               "_lagsar", "_errorsar", "_fcor"), specials = c("residuals", 
               "rescor")), class = c("brmsfamily", "family")), mecor = TRUE), class = c("brmsformula", 
    ...
3: do.call("get_contrasts", args)
2: comparisons(model, newdata = newdata, variables = variables, 
       vcov = vcov, conf_level = conf_level, type = type, wts = wts, 
       hypothesis = hypothesis, equivalence = equivalence, df = df, 
       p_adjust = p_adjust, by = by, eps = eps, numderiv = numderiv, 
       comparison = slope, cross = FALSE, internal_call = TRUE, 
       ...)
1: slopes(mod_miss, newdata = datagrid(woman = 1, passengerClass = "1st"))

#--------------------------------------
## Example 3
7: stop(format_message(string = string, ..., line_length = line_length, 
       indent = indent), call. = call)
6: format_alert(..., type = "error")
5: insight::format_error(msg)
4: sanitize_variables(model = model, newdata = newdata, modeldata = modeldata, 
       variables = variables, cross = cross, by = by, comparison = comparison, 
       eps = eps)
3: comparisons(model, newdata = newdata, variables = variables, 
       vcov = vcov, conf_level = conf_level, type = type, wts = wts, 
       hypothesis = hypothesis, equivalence = equivalence, df = df, 
       p_adjust = p_adjust, by = by, eps = eps, numderiv = numderiv, 
       comparison = slope, cross = FALSE, internal_call = TRUE, 
       ...)
2: slopes(mod_miss, variables = "age", newdata = datagrid(woman = 0:1))
1: posterior_draws(slopes(mod_miss, variables = "age", newdata = datagrid(woman = 0:1)))

sessionInfo()

R version 4.3.0 (2023-04-21 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19045)

Matrix products: default


locale:
[1] LC_COLLATE=English_South Africa.utf8  LC_CTYPE=English_South Africa.utf8    LC_MONETARY=English_South Africa.utf8
[4] LC_NUMERIC=C                          LC_TIME=English_South Africa.utf8    

time zone: Africa/Johannesburg
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] brms_2.21.0            Rcpp_1.0.12            marginaleffects_0.21.0

loaded via a namespace (and not attached):
 [1] bridgesampling_1.1-2 tensorA_0.36.2.1     utf8_1.2.4           generics_0.1.3       stringi_1.8.4       
 [6] lattice_0.21-8       magrittr_2.0.3       grid_4.3.0           mvtnorm_1.2-5        jsonlite_1.8.8      
[11] Matrix_1.5-4         processx_3.8.4       pkgbuild_1.4.4       backports_1.5.0      ps_1.7.7            
[16] gridExtra_2.3        Brobdingnag_1.2-9    fansi_1.0.6          QuickJSR_1.2.2       scales_1.3.0        
[21] codetools_0.2-19     abind_1.4-5          cli_3.6.3            rlang_1.1.4          cmdstanr_0.8.1      
[26] munsell_0.5.1        withr_3.0.0          StanHeaders_2.32.9   tools_4.3.0          rstan_2.32.6        
[31] inline_0.3.19        parallel_4.3.0       rstantools_2.4.0     checkmate_2.3.1      coda_0.19-4.1       
[36] dplyr_1.1.4          colorspace_2.1-0     ggplot2_3.5.1        vctrs_0.6.5          posterior_1.6.0     
[41] R6_2.5.1             matrixStats_1.3.0    stats4_4.3.0         lifecycle_1.0.4      stringr_1.5.1       
[46] insight_0.20.1.13    pkgconfig_2.0.3      RcppParallel_5.1.7   pillar_1.9.0         gtable_0.3.5        
[51] loo_2.8.0            data.table_1.15.4    glue_1.7.0           collapse_2.0.15      xfun_0.45           
[56] tibble_3.2.1         tidyselect_1.2.1     knitr_1.47           rstudioapi_0.16.0    bayesplot_1.11.1    
[61] nlme_3.1-162         compiler_4.3.0       distributional_0.4.0

[strengejacke]

Maybe related to #896? Not sure, will check later.