Informative prior archetypes

Informative prior archetypes allow users to conveniently set informative priors in brms.mmrm in a robust way, guarding against common pitfalls such as reference level issues, interpretation problems, and rank deficiency.

Constructing an archetype

We begin with the FEV dataset from the mmrm package, an artificial (simulated) dataset of a clinical trial investigating the effect of an active treatment on FEV1 (forced expired volume in one second), compared to placebo. FEV1 is a measure of how quickly the lungs can be emptied and low levels may indicate chronic obstructive pulmonary disease (COPD).

The dataset is a tibble with 800 rows and 7 variables:

We will derive FEV1_CHG = FEV1 - FEV1_BL and analyze FEV1_CHG as the outcome variable.

library(brms.mmrm)
data(fev_data, package = "mmrm")
data <- fev_data |>
  mutate("FEV1_CHG" = FEV1 - FEV1_BL) |>
  brm_data(
    outcome = "FEV1_CHG",
    role = "change",
    group = "ARMCD",
    time = "AVISIT",
    patient = "USUBJID",
    baseline = "FEV1_BL",
    reference_group = "PBO",
    covariates = c("WEIGHT", "SEX")
  )
data
#> # A tibble: 800 × 7
#>    FEV1_CHG FEV1_BL ARMCD AVISIT USUBJID WEIGHT SEX   
#>       <dbl>   <dbl> <chr> <chr>  <fct>    <dbl> <fct> 
#>  1    NA       45.0 PBO   VIS1   PT2      0.465 Male  
#>  2   -13.6     45.0 PBO   VIS2   PT2      0.233 Male  
#>  3    -8.15    45.0 PBO   VIS3   PT2      0.360 Male  
#>  4     3.78    45.0 PBO   VIS4   PT2      0.507 Male  
#>  5    NA       43.5 PBO   VIS1   PT3      0.682 Female
#>  6    -7.51    43.5 PBO   VIS2   PT3      0.892 Female
#>  7    NA       43.5 PBO   VIS3   PT3      0.128 Female
#>  8    -6.34    43.5 PBO   VIS4   PT3      0.222 Female
#>  9   -11.3     43.6 PBO   VIS1   PT5      0.411 Male  
#> 10    NA       43.6 PBO   VIS2   PT5      0.422 Male  
#> # ℹ 790 more rows

The functions listed at https://openpharma.github.io/brms.mmrm/reference/index.html#informative-prior-archetypes can create different kinds of informative prior archetypes from a dataset like the one above. For example, suppose we want to place informative priors on the successive differences between adjacent time points. This approach is appropriate and desirable in many situations because the structure naturally captures the prior correlations among adjacent visits of a clinical trial. To do this, we create an instance of the “successive cells” archetype.

archetype <- brm_archetype_successive_cells(data, baseline = FALSE)

The instance of the archetype is an ordinary tibble, but it adds new columns.

archetype
#> # A tibble: 800 × 21
#>    x_PBO_VIS1 x_PBO_VIS2 x_PBO_VIS3 x_PBO_VIS4 x_TRT_VIS1 x_TRT_VIS2 x_TRT_VIS3 x_TRT_VIS4
#>  *      <dbl>      <dbl>      <dbl>      <dbl>      <dbl>      <dbl>      <dbl>      <dbl>
#>  1          1          0          0          0          0          0          0          0
#>  2          1          1          0          0          0          0          0          0
#>  3          1          1          1          0          0          0          0          0
#>  4          1          1          1          1          0          0          0          0
#>  5          1          0          0          0          0          0          0          0
#>  6          1          1          0          0          0          0          0          0
#>  7          1          1          1          0          0          0          0          0
#>  8          1          1          1          1          0          0          0          0
#>  9          1          0          0          0          0          0          0          0
#> 10          1          1          0          0          0          0          0          0
#> # ℹ 790 more rows
#> # ℹ 13 more variables: nuisance_WEIGHT <dbl>, nuisance_SEX_Male <dbl>,
#> #   nuisance_FEV1_BL.AVISITVIS1 <dbl>, nuisance_FEV1_BL.AVISITVIS2 <dbl>,
#> #   nuisance_FEV1_BL.AVISITVIS3 <dbl>, nuisance_FEV1_BL.AVISITVIS4 <dbl>, USUBJID <fct>,
#> #   AVISIT <chr>, FEV1_CHG <dbl>, FEV1_BL <dbl>, ARMCD <chr>, WEIGHT <dbl>, SEX <fct>

Those new columns constitute a custom model matrix to describe the desired parameterization. We have effects of interest to express successive differences:

attr(archetype, "brm_archetype_interest")
#> [1] "x_PBO_VIS1" "x_PBO_VIS2" "x_PBO_VIS3" "x_PBO_VIS4" "x_TRT_VIS1" "x_TRT_VIS2" "x_TRT_VIS3"
#> [8] "x_TRT_VIS4"

In addition, we have nuisance variables. Some nuisance variables are continuous covariates, while others are levels of one-hot-encoded concomitant factors or interactions of those concomitant factors with baseline and/or subgroup. All nuisance variables are centered at their means so the reference level of the model is at the “center” of the data and not implicitly conditional on a subset of the data.1 In addition, some nuisance variables are automatically dropped in order to ensure the model matrix is full-rank. This is critically important to preserve the interpretation of the columns of interest and make sure the informative priors behave as expected.

attr(archetype, "brm_archetype_nuisance")
#> [1] "nuisance_WEIGHT"             "nuisance_SEX_Male"           "nuisance_FEV1_BL.AVISITVIS1"
#> [4] "nuisance_FEV1_BL.AVISITVIS2" "nuisance_FEV1_BL.AVISITVIS3" "nuisance_FEV1_BL.AVISITVIS4"

The factors of interest linearly map to marginal means. To see the mapping, call summary() on the archetype. The printed output helps build intuition on how the archetype is parameterized and what those parameters are doing.

summary(archetype)
#> # This is the "successive cells" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = x_PBO_VIS1
#> #    PBO:VIS2 = x_PBO_VIS1 + x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_PBO_VIS4
#> #    TRT:VIS1 = x_TRT_VIS1
#> #    TRT:VIS2 = x_TRT_VIS1 + x_TRT_VIS2
#> #    TRT:VIS3 = x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3
#> #    TRT:VIS4 = x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3 + x_TRT_VIS4

Informative priors

Let’s assume you want to assign informative priors to the fixed effect parameters of interest declared in the archetype, such as x_group_1_time_2 and x_group_2_time_3. Your priors may come from expert elicitation, historical data, or some other method, and you might consider distributional families recommended by the Stan team. However you construct these priors, brms.mmrm helps you assign them to the model without having to guess at the automatically-generated names of model coefficients in R.

In the printed output from summary(archetype), parameters of interest such as x_group_1_time_2 and x_group_2_time_3 are always labeled using treatment groups and time points in the data (and subgroup levels, if applicable). This labeling mechanism is the same regardless of which archetype you choose, and it the way brms.mmrm helps you assign priors.

brm_prior_label() is one way to create a labeling scheme. Each call to brm_prior_label() below assigns a univariate prior to a fixed effect parameter. Each univariate prior is a Stan code string. Possible choices are documented in the Stan function reference at https://mc-stan.org/docs/functions-reference/unbounded_continuous_distributions.html.

label <- NULL |>
  brm_prior_label(code = "student_t(4, -7.57, 4.96)", group = "PBO", time = "VIS1") |>
  brm_prior_label(code = "student_t(4,  3.14, 7.86)", group = "PBO", time = "VIS2") |>
  brm_prior_label(code = "student_t(4,  8.78, 8.18)", group = "PBO", time = "VIS3") |>
  brm_prior_label(code = "student_t(4,  3.36, 8.10)", group = "PBO", time = "VIS4") |>
  brm_prior_label(code = "student_t(4, -2.96, 4.78)", group = "TRT", time = "VIS1") |>
  brm_prior_label(code = "student_t(4,  3.13, 7.64)", group = "TRT", time = "VIS2") |>
  brm_prior_label(code = "student_t(4,  7.65, 8.24)", group = "TRT", time = "VIS3") |>
  brm_prior_label(code = "student_t(4,  4.64, 8.21)", group = "TRT", time = "VIS4")
label
#> # A tibble: 8 × 3
#>   code                      group time 
#>   <chr>                     <chr> <chr>
#> 1 student_t(4, -7.57, 4.96) PBO   VIS1 
#> 2 student_t(4,  3.14, 7.86) PBO   VIS2 
#> 3 student_t(4,  8.78, 8.18) PBO   VIS3 
#> 4 student_t(4,  3.36, 8.10) PBO   VIS4 
#> 5 student_t(4, -2.96, 4.78) TRT   VIS1 
#> 6 student_t(4,  3.13, 7.64) TRT   VIS2 
#> 7 student_t(4,  7.65, 8.24) TRT   VIS3 
#> 8 student_t(4,  4.64, 8.21) TRT   VIS4

As an alternative to brm_prior_label(), you can start with a template and manually fill in the Stan code.

template <- brm_prior_template(archetype)
template
#> # A tibble: 8 × 3
#>   code  group time 
#>   <chr> <chr> <chr>
#> 1 <NA>  PBO   VIS1 
#> 2 <NA>  PBO   VIS2 
#> 3 <NA>  PBO   VIS3 
#> 4 <NA>  PBO   VIS4 
#> 5 <NA>  TRT   VIS1 
#> 6 <NA>  TRT   VIS2 
#> 7 <NA>  TRT   VIS3 
#> 8 <NA>  TRT   VIS4
label <- template |>
  mutate(
    code = c(
      "student_t(4, -7.57, 4.96)",
      "student_t(4,  3.14, 7.86)",
      "student_t(4,  8.78, 8.18)",
      "student_t(4,  3.36, 8.10)",
      "student_t(4, -2.96, 4.78)",
      "student_t(4,  3.13, 7.64)",
      "student_t(4,  7.65, 8.24)",
      "student_t(4,  4.64, 8.21)"
    )
  )
label
#> # A tibble: 8 × 3
#>   code                      group time 
#>   <chr>                     <chr> <chr>
#> 1 student_t(4, -7.57, 4.96) PBO   VIS1 
#> 2 student_t(4,  3.14, 7.86) PBO   VIS2 
#> 3 student_t(4,  8.78, 8.18) PBO   VIS3 
#> 4 student_t(4,  3.36, 8.10) PBO   VIS4 
#> 5 student_t(4, -2.96, 4.78) TRT   VIS1 
#> 6 student_t(4,  3.13, 7.64) TRT   VIS2 
#> 7 student_t(4,  7.65, 8.24) TRT   VIS3 
#> 8 student_t(4,  4.64, 8.21) TRT   VIS4

After you have a labeling scheme, brm_prior_archetype() can create a brms prior for the important fixed effects.2

prior <- brm_prior_archetype(label = label, archetype = archetype)
prior
#>                      prior class       coef group resp dpar nlpar   lb   ub source
#>  student_t(4, -7.57, 4.96)     b x_PBO_VIS1                       <NA> <NA>   user
#>  student_t(4,  3.14, 7.86)     b x_PBO_VIS2                       <NA> <NA>   user
#>  student_t(4,  8.78, 8.18)     b x_PBO_VIS3                       <NA> <NA>   user
#>  student_t(4,  3.36, 8.10)     b x_PBO_VIS4                       <NA> <NA>   user
#>  student_t(4, -2.96, 4.78)     b x_TRT_VIS1                       <NA> <NA>   user
#>  student_t(4,  3.13, 7.64)     b x_TRT_VIS2                       <NA> <NA>   user
#>  student_t(4,  7.65, 8.24)     b x_TRT_VIS3                       <NA> <NA>   user
#>  student_t(4,  4.64, 8.21)     b x_TRT_VIS4                       <NA> <NA>   user

In less common situations, you may wish to assign priors to nuisance parameters. For example, our model accounts for interactions between baseline and discrete time, and it may be reasonable to assign priors to these slopes based on high-quality historical data. This requires a thorough understanding of the fixed effect structure of the model, but it can be done directly through brms. First, check the formula for the included nuisance parameters. brm_formula() automatically understands archetypes.

brm_formula(archetype)
#> FEV1_CHG ~ 0 + x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_PBO_VIS4 + x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3 + x_TRT_VIS4 + nuisance_WEIGHT + nuisance_SEX_Male + nuisance_FEV1_BL.AVISITVIS1 + nuisance_FEV1_BL.AVISITVIS2 + nuisance_FEV1_BL.AVISITVIS3 + nuisance_FEV1_BL.AVISITVIS4 + unstr(time = AVISIT, gr = USUBJID) 
#> sigma ~ 0 + AVISIT

The "nuisance_*" terms are the nuisance variables, and the ones involving baseline are nuisance_FEV1_BL.AVISITVIS1, nuisance_FEV1_BL.AVISITVIS2, nuisance_FEV1_BL.AVISITVIS3, and nuisance_FEV1_BL.AVISITVIS4. Because there is no overall slope for baseline, we can interpret each term as the linear rate of change in the outcome variable per unit increase in baseline for a given discrete time point. Suppose we use this interpretation to construct informative priors student_t(4, -0.83, 1), student_t(4, -0.78, 1), student_t(4, -0.86, 1), and student_t(4, -0.82, 1), respectively. Use brms::set_prior() and c() to append these priors to our existing prior object:

prior <- c(
  prior,
  set_prior("student_t(4, -0.83, 1)", coef = "nuisance_FEV1_BL.AVISITVIS1"),
  set_prior("student_t(4, -0.78, 1)", coef = "nuisance_FEV1_BL.AVISITVIS2"),
  set_prior("student_t(4, -0.86, 1)", coef = "nuisance_FEV1_BL.AVISITVIS3"),
  set_prior("student_t(4, -0.82, 1)", coef = "nuisance_FEV1_BL.AVISITVIS4")
)
prior
#>                      prior class                        coef group resp dpar nlpar   lb   ub
#>  student_t(4, -7.57, 4.96)     b                  x_PBO_VIS1                       <NA> <NA>
#>  student_t(4,  3.14, 7.86)     b                  x_PBO_VIS2                       <NA> <NA>
#>  student_t(4,  8.78, 8.18)     b                  x_PBO_VIS3                       <NA> <NA>
#>  student_t(4,  3.36, 8.10)     b                  x_PBO_VIS4                       <NA> <NA>
#>  student_t(4, -2.96, 4.78)     b                  x_TRT_VIS1                       <NA> <NA>
#>  student_t(4,  3.13, 7.64)     b                  x_TRT_VIS2                       <NA> <NA>
#>  student_t(4,  7.65, 8.24)     b                  x_TRT_VIS3                       <NA> <NA>
#>  student_t(4,  4.64, 8.21)     b                  x_TRT_VIS4                       <NA> <NA>
#>     student_t(4, -0.83, 1)     b nuisance_FEV1_BL.AVISITVIS1                       <NA> <NA>
#>     student_t(4, -0.78, 1)     b nuisance_FEV1_BL.AVISITVIS2                       <NA> <NA>
#>     student_t(4, -0.86, 1)     b nuisance_FEV1_BL.AVISITVIS3                       <NA> <NA>
#>     student_t(4, -0.82, 1)     b nuisance_FEV1_BL.AVISITVIS4                       <NA> <NA>
#>  source
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user
#>    user

The model still has many parameters where we did not set priors, and brms sets automatic defaults. You can see these defaults with brms::get_prior().

brms::get_prior(formula = formula, data = archetype)
#> Warning: Rows containing NAs were excluded from the model.
#>   prior   class                        coef group resp  dpar nlpar lb ub       source
#>  (flat)       b                                                               default
#>  (flat)       b nuisance_FEV1_BL.AVISITVIS1                              (vectorized)
#>  (flat)       b nuisance_FEV1_BL.AVISITVIS2                              (vectorized)
#>  (flat)       b nuisance_FEV1_BL.AVISITVIS3                              (vectorized)
#>  (flat)       b nuisance_FEV1_BL.AVISITVIS4                              (vectorized)
#>  (flat)       b           nuisance_SEX_Male                              (vectorized)
#>  (flat)       b             nuisance_WEIGHT                              (vectorized)
#>  (flat)       b                  x_PBO_VIS1                              (vectorized)
#>  (flat)       b                  x_PBO_VIS2                              (vectorized)
#>  (flat)       b                  x_PBO_VIS3                              (vectorized)
#>  (flat)       b                  x_PBO_VIS4                              (vectorized)
#>  (flat)       b                  x_TRT_VIS1                              (vectorized)
#>  (flat)       b                  x_TRT_VIS2                              (vectorized)
#>  (flat)       b                  x_TRT_VIS3                              (vectorized)
#>  (flat)       b                  x_TRT_VIS4                              (vectorized)
#>  lkj(1) cortime                                                               default
#>  (flat)       b                                        sigma                  default
#>  (flat)       b                  AVISITVIS1            sigma             (vectorized)
#>  (flat)       b                  AVISITVIS2            sigma             (vectorized)
#>  (flat)       b                  AVISITVIS3            sigma             (vectorized)
#>  (flat)       b                  AVISITVIS4            sigma             (vectorized)

https://paul-buerkner.github.io/brms/reference/set_prior.html documents many of the default priors set by brms. In particular, "(flat)" denotes an improper uniform prior over all the real numbers.

Modeling and analysis

The downstream methods in brms.mmrm automatically understand how to work with informative prior archetypes. Notably, the formula uses custom interest and nuisance variables instead of the original variables in the data.

formula <- brm_formula(archetype)
formula
#> FEV1_CHG ~ 0 + x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_PBO_VIS4 + x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3 + x_TRT_VIS4 + nuisance_WEIGHT + nuisance_SEX_Male + nuisance_FEV1_BL.AVISITVIS1 + nuisance_FEV1_BL.AVISITVIS2 + nuisance_FEV1_BL.AVISITVIS3 + nuisance_FEV1_BL.AVISITVIS4 + unstr(time = AVISIT, gr = USUBJID) 
#> sigma ~ 0 + AVISIT

The model can accept the archetype, formula, and prior. Usage is the same as in non-archetype workflows.

model <- brm_model(
  data = archetype,
  formula = formula,
  prior = prior,
  refresh = 0
)
#> Compiling Stan program...
#> Start sampling
brms::prior_summary(model)
#>                      prior    class                        coef group resp  dpar nlpar lb ub
#>                     (flat)        b                                                         
#>     student_t(4, -0.83, 1)        b nuisance_FEV1_BL.AVISITVIS1                             
#>     student_t(4, -0.78, 1)        b nuisance_FEV1_BL.AVISITVIS2                             
#>     student_t(4, -0.86, 1)        b nuisance_FEV1_BL.AVISITVIS3                             
#>     student_t(4, -0.82, 1)        b nuisance_FEV1_BL.AVISITVIS4                             
#>                     (flat)        b           nuisance_SEX_Male                             
#>                     (flat)        b             nuisance_WEIGHT                             
#>  student_t(4, -7.57, 4.96)        b                  x_PBO_VIS1                             
#>  student_t(4,  3.14, 7.86)        b                  x_PBO_VIS2                             
#>  student_t(4,  8.78, 8.18)        b                  x_PBO_VIS3                             
#>  student_t(4,  3.36, 8.10)        b                  x_PBO_VIS4                             
#>  student_t(4, -2.96, 4.78)        b                  x_TRT_VIS1                             
#>  student_t(4,  3.13, 7.64)        b                  x_TRT_VIS2                             
#>  student_t(4,  7.65, 8.24)        b                  x_TRT_VIS3                             
#>  student_t(4,  4.64, 8.21)        b                  x_TRT_VIS4                             
#>                     (flat)        b                                        sigma            
#>                     (flat)        b                  AVISITVIS1            sigma            
#>                     (flat)        b                  AVISITVIS2            sigma            
#>                     (flat)        b                  AVISITVIS3            sigma            
#>                     (flat)        b                  AVISITVIS4            sigma            
#>       lkj_corr_cholesky(1) Lcortime                                                         
#>        source
#>       default
#>          user
#>          user
#>          user
#>          user
#>  (vectorized)
#>  (vectorized)
#>          user
#>          user
#>          user
#>          user
#>          user
#>          user
#>          user
#>          user
#>       default
#>  (vectorized)
#>  (vectorized)
#>  (vectorized)
#>  (vectorized)
#>       default

Marginal mean estimation, post-processing, and visualization automatically understand the archetype without any user intervention.

draws <- brm_marginal_draws(
  data = archetype,
  formula = formula,
  model = model
)
summaries_model <- brm_marginal_summaries(draws)
summaries_data <- brm_marginal_data(archetype)
brm_plot_compare(model = summaries_model, data = summaries_data)
plot of chunk archetype_compare_data

plot of chunk archetype_compare_data

Other archetypes

Other informative prior archetypes use different fixed effects. For example, brms.mmrm supports simple cell mean and treatment effect parameterizations.

summary(brm_archetype_cells(data))
#> # This is the "cells" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = x_PBO_VIS1
#> #    PBO:VIS2 = x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS4
#> #    TRT:VIS1 = x_TRT_VIS1
#> #    TRT:VIS2 = x_TRT_VIS2
#> #    TRT:VIS3 = x_TRT_VIS3
#> #    TRT:VIS4 = x_TRT_VIS4
summary(brm_archetype_effects(data))
#> # This is the "effects" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = x_PBO_VIS1
#> #    PBO:VIS2 = x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS4
#> #    TRT:VIS1 = x_PBO_VIS1 + x_TRT_VIS1
#> #    TRT:VIS2 = x_PBO_VIS2 + x_TRT_VIS2
#> #    TRT:VIS3 = x_PBO_VIS3 + x_TRT_VIS3
#> #    TRT:VIS4 = x_PBO_VIS4 + x_TRT_VIS4

There are archetypes to parameterize the average across all time points in the data. Below, x_group_1_time_2 is the average across time points for group 1 because it is the algebraic result of simplifying (group_1:time_2 + group_1:time_3 + group_1:time_3) / 3.

summary(brm_archetype_average_cells(data))
#> # This is the "average cells" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = 4*x_PBO_VIS1 - x_PBO_VIS2 - x_PBO_VIS3 - x_PBO_VIS4
#> #    PBO:VIS2 = x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS4
#> #    TRT:VIS1 = 4*x_TRT_VIS1 - x_TRT_VIS2 - x_TRT_VIS3 - x_TRT_VIS4
#> #    TRT:VIS2 = x_TRT_VIS2
#> #    TRT:VIS3 = x_TRT_VIS3
#> #    TRT:VIS4 = x_TRT_VIS4

There is also a treatment effect version where x_group_2_time_2 becomes the time-averaged treatment effect of group 2 relative to group 1.

summary(brm_archetype_average_effects(data))
#> # This is the "average effects" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = 4*x_PBO_VIS1 - x_PBO_VIS2 - x_PBO_VIS3 - x_PBO_VIS4
#> #    PBO:VIS2 = x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS4
#> #    TRT:VIS1 = 4*x_PBO_VIS1 - x_PBO_VIS2 - x_PBO_VIS3 - x_PBO_VIS4 + 4*x_TRT_VIS1 - x_TRT_VIS2 - x_TRT_VIS3 - x_TRT_VIS4
#> #    TRT:VIS2 = x_PBO_VIS2 + x_TRT_VIS2
#> #    TRT:VIS3 = x_PBO_VIS3 + x_TRT_VIS3
#> #    TRT:VIS4 = x_PBO_VIS4 + x_TRT_VIS4

In addition, there is a treatment effect version of the successive differences archetype from earlier in the vignette.

summary(brm_archetype_successive_effects(data))
#> # This is the "successive effects" informative prior archetype in brms.mmrm.
#> # The following equations show the relationships between the
#> # marginal means (left-hand side) and fixed effect parameters
#> # (right-hand side). You can create informative priors for the
#> # fixed effect parameters using historical borrowing,
#> # expert elicitation, or other methods.
#> # 
#> #    PBO:VIS1 = x_PBO_VIS1
#> #    PBO:VIS2 = x_PBO_VIS1 + x_PBO_VIS2
#> #    PBO:VIS3 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3
#> #    PBO:VIS4 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_PBO_VIS4
#> #    TRT:VIS1 = x_PBO_VIS1 + x_TRT_VIS1
#> #    TRT:VIS2 = x_PBO_VIS1 + x_PBO_VIS2 + x_TRT_VIS1 + x_TRT_VIS2
#> #    TRT:VIS3 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3
#> #    TRT:VIS4 = x_PBO_VIS1 + x_PBO_VIS2 + x_PBO_VIS3 + x_PBO_VIS4 + x_TRT_VIS1 + x_TRT_VIS2 + x_TRT_VIS3 + x_TRT_VIS4

  1. brm_recenter_nuisance() can retroactively recenter a nuisance column to a fixed value other than its mean.↩︎

  2. brms priors are documented in https://paul-buerkner.github.io/brms/reference/set_prior.html.↩︎