brms-rstan: re-organize static files

Author: falkmielke

content/tutorials/r_brms/brms_eng/bayesian_statistics_1_script_eng.R

@@ -48,7 +48,7 @@ confint(lm1, level = 0.9)
 
 
 ## -------------------------------------------------------------------------------------------------------------------------------------
-source(file = "./source/mcmc_functions.R")
+source(file = here("code", "brms_modeling", "mcmc_functions.R"))
 
 
 ## -------------------------------------------------------------------------------------------------------------------------------------
@@ -230,7 +230,7 @@ fit_normal1 <- brm(
   family = gaussian(),         # we use the Normal distribution
   data = ants_df,              # specify data
   chains = nchains,            # MCMC parameters
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -335,7 +335,7 @@ mcmc_neff(ratios_fit_normal1) + yaxis_text(hjust = 1)
 
 ## ----simple-model-fit1----------------------------------------------------------------------------------------------------------------
 # Visualise model fit via bayesplot package
-pp_check(fit_normal1, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_normal1, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -346,7 +346,7 @@ fit_poisson1 <- brm(
   family = poisson(),          # we use the Poisson distribution
   data = ants_df,              # specify the data
   chains = nchains,            # MCMC parameters
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -366,7 +366,7 @@ mcmc_rhat(rhats_fit_poisson1) + yaxis_text(hjust = 1)
 
 ## ----poisson-model-fit-vis------------------------------------------------------------------------------------------------------------
 # Visualise model fit via bayesplot package
-pp_check(fit_poisson1, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_poisson1, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -377,7 +377,7 @@ fit_poisson2 <- brm(
   family = poisson(),
   data = ants_df,
   chains = nchains,
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -399,7 +399,7 @@ mcmc_rhat(rhats_fit_poisson2) + yaxis_text(hjust = 1)
 ## ----rand-intercept-model-fit-vis-----------------------------------------------------------------------------------------------------
 # Visualise model fit of the Poisson model with random intercept via
 # bayesplot package
-pp_check(fit_poisson2, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_poisson2, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -518,7 +518,7 @@ fit_poisson2 %>%
   # calculate average numbers and convert to long format for visualisation
   mutate(bog = exp(b_Intercept),
          forest = exp(b_Intercept + b_habitatForest)) %>%
-  pivot_longer(cols = c("bog", "forest"), names_to = "habitat", 
+  pivot_longer(cols = c("bog", "forest"), names_to = "habitat",
                values_to = "sp_rich") %>%
   # visualise via ggplot()
   ggplot(aes(y = sp_rich, x = habitat)) +

content/tutorials/r_brms/brms_eng/workshop_1_mcmc_en_brms_eng.Rmd

@@ -298,7 +298,7 @@ confint(lm1, level = 0.9)
 We source some short functions to calculate the (log) likelihood and the prior and to execute the MCMC metropolis algorithm.
 
 ```{r}
-source(file = "./mcmc_functions.R")
+source(file = here("code", "brms_modeling", "mcmc_functions.R"))
 ```
 
 For this simple model with a small data set, we can calculate and plot the posterior for a large number of combinations of 'beta_0' and 'beta_1'.
@@ -1219,7 +1219,7 @@ conflicted::conflicts_prefer(brms::loo)
 ```
 
 ## Model Definition
-RMarkdown can handle `stan` code chunks, though more general model definition is outsourced to a separate "*.stan" file.
+RMarkdown can handle `stan` code chunks, though more general model definition is outsourced to a separate "*.stan" file (e.g. [here](https://github.com/inbo/tutorials/tree/master/static/code/brms_modeling/poisson_model.stan)).
 Alternatively, you can define your model in a big text block, as shown below.
 The simple poisson model resembles [one of the `stan`-dard examples](https://mc-stan.org/docs/stan-users-guide/posterior-prediction.html#posterior-prediction-for-regressions), which you can refer to for all further details and more.
 
@@ -1260,7 +1260,7 @@ When working outside RStudio/RMarkdown, you might prefer loading the model from
 
 ```{r stan_load_model, eval=TRUE, class.source='fold-show'}
 stan_poisson_model <- stan_model(
-  # file = "./poisson_model.stan",
+  # file = here("code", "brms_modeling", "poisson_model.stan"),
   model_code = stan_poisson_model_code, 
   model_name = "stan poisson model"
   )

content/tutorials/r_brms/brms_nl/bayesian_statistics_1_script.R

@@ -48,7 +48,7 @@ confint(lm1, level = 0.9)
 
 
 ## -------------------------------------------------------------------------------------------------------------------------------------
-source(file = "./source/mcmc_functions.R")
+source(file = here::here("code", "brms_modeling", "mcmc_functions.R"))
 
 
 ## -------------------------------------------------------------------------------------------------------------------------------------
@@ -232,7 +232,7 @@ fit_normal1 <- brm(
   family = gaussian(),         # we gebruiken de Normaal verdeling
   data = ants_df,              # ingeven data
   chains = nchains,            # MCMC parameters
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -298,7 +298,7 @@ as_draws_df(fit_normal1, variable = parameters) %>%
 
 
 ## ----simpel-model-posterior-density2--------------------------------------------------------------------------------------------------
-# Visualisatie density plot van de posterior voor ieder van de chains apart in 
+# Visualisatie density plot van de posterior voor ieder van de chains apart in
 # overlay. via Bayesplot package
 mcmc_dens_overlay(fit_normal1, pars = parameters)
 
@@ -337,7 +337,7 @@ mcmc_neff(ratios_fit_normal1) + yaxis_text(hjust = 1)
 
 ## ----simpel-model-fit1----------------------------------------------------------------------------------------------------------------
 # Visualiseer model fit via Bayesplot package
-pp_check(fit_normal1, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_normal1, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -348,7 +348,7 @@ fit_poisson1 <- brm(
   family = poisson(),          # we gebruiken de Poisson verdeling
   data = ants_df,              # ingeven data
   chains = nchains,            # MCMC parameters
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -368,7 +368,7 @@ mcmc_rhat(rhats_fit_poisson1) + yaxis_text(hjust = 1)
 
 ## ----poisson-model-fit-vis------------------------------------------------------------------------------------------------------------
 # Visualiseer model fit via Bayesplot package
-pp_check(fit_poisson1, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_poisson1, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -379,7 +379,7 @@ fit_poisson2 <- brm(
   family = poisson(),
   data = ants_df,
   chains = nchains,
-  warmup = burnin, 
+  warmup = burnin,
   iter = niter,
   cores = nparallel,
   thin = thinning,
@@ -401,7 +401,7 @@ mcmc_rhat(rhats_fit_poisson2) + yaxis_text(hjust = 1)
 ## ----rand-intercept-model-fit-vis-----------------------------------------------------------------------------------------------------
 # Visualiseer model fit van het Poisson model met random intercept via
 # Bayesplot package
-pp_check(fit_poisson2, type = "dens_overlay_grouped", ndraws = 100, 
+pp_check(fit_poisson2, type = "dens_overlay_grouped", ndraws = 100,
          group = "habitat")
 
 
@@ -520,7 +520,7 @@ fit_poisson2 %>%
   # bereken gemiddelde aantallen en zet om naar lang formaat voor visualisatie
   mutate(bog = exp(b_Intercept),
          forest = exp(b_Intercept + b_habitatForest)) %>%
-  pivot_longer(cols = c("bog", "forest"), names_to = "habitat", 
+  pivot_longer(cols = c("bog", "forest"), names_to = "habitat",
                values_to = "sp_rich") %>%
   # visualiseer via ggplot()
   ggplot(aes(y = sp_rich, x = habitat)) +

content/tutorials/r_brms/brms_nl/install_packages.R

@@ -294,7 +294,8 @@ confint(lm1, level = 0.9)
 We lezen enkele korte functies in om de (log) likelihood en de prior te berekenen en het MCMC metropolis algoritme uit te voeren.
 
 ```{r}
-source(file = "./mcmc_functions.R")
+
+source(file = here("code", "brms_modeling", "mcmc_functions.R"))
 ```
 
 Voor dit eenvoudig model met een kleine dataset kunnen we de posterior voor een groot aantal combinaties voor `beta_0` en `beta_1` uitrekenen en plotten.
@@ -1206,7 +1207,7 @@ conflicted::conflicts_prefer(brms::loo)
 ```
 
 ## Model Definition
-RMarkdown kan `stan`-chunks verwerken, hoewel de algemenere modeldefinitie is uitbesteed aan een apart "*.stan"-bestand.
+RMarkdown kan `stan`-chunks verwerken, hoewel de algemenere modeldefinitie is uitbesteed aan een apart "*.stan"-bestand (bv. [hier](https://github.com/inbo/tutorials/tree/master/static/code/brms_modeling/poisson_model.stan)).
 Alternatief kan je je model ook als een groot tekstblok definiëren, zoals hieronder te zien.
 Het eenvoudige poissonmodel lijkt op [een van de `stan`-daardvoorbeelden](https://mc-stan.org/docs/stan-users-guide/posterior-prediction.html#posterior-prediction-for-regressions), waarnaar u terecht kunt voor verdere details en meer.
 
@@ -1247,7 +1248,7 @@ Als je buiten RStudio/RMarkdown werkt, kun je het model het beste laden vanuit e
 
 ```{r stan_load_model, eval=TRUE, class.source='fold-show'}
 stan_poisson_model <- stan_model(
-  # file = "./poisson_model.stan",
+  # file = here("code", "brms_modeling", "poisson_model.stan"),
   model_code = stan_poisson_model_code, 
   model_name = "stan poisson model"
   )

content/tutorials/r_brms/brms_nl/mcmc_functions.R

@@ -8,10 +8,11 @@ tags: ["generalized linear regression", "brms", "r", "mixed models", "stan"]
 ---
 
 In the fall of 2023, a tutorial on Bayesian statistics with the [**brms**](https://paul-buerkner.github.io/brms/) packages was organised at INBO.
-Before you start the tutorial, please follow the instructions in [this R script](https://github.com/inbo/tutorials/blob/master/content/tutorials/r_brms/brms_eng/install_packages.R) to install brms properly.
+Before you start the tutorial, please follow the instructions in [this R script](https://github.com/inbo/tutorials/blob/master/static/code/brms_modeling/install_packages.R) to install brms properly.
 For more advanced use cases, we added an example of how to directly use [**stan**](https://mc-stan.org) via the `rstan` package.
 All course material can be found in Dutch and English on the following pages:
 
 - The [English tutorial](../../html/workshop_1_mcmc_en_brms_eng.html) with all [course material](https://github.com/inbo/tutorials/tree/master/content/tutorials/r_brms/brms_eng) to test the examples yourself.
 - The [Dutch tutorial](../../html/workshop_1_mcmc_en_brms.html) with all [course material](https://github.com/inbo/tutorials/tree/master/content/tutorials/r_brms/brms_nl) to test the examples yourself.
+- [General files](https://github.com/inbo/tutorials/tree/master/static/code/brms_modeling) useful for both tutorials.