Getting bayesian anova modules up to speed with changes in master

data-raw/bf_anova_precalculations.r

@@ -24,8 +24,8 @@ future::plan(multisession, workers = n_cores)
 # Calculation configurations ---------------------------------------------------
 
 # Dataframe is created with "./data-raw/options/anova_options_bayesian.R
-bayes_anova_options <- readr::read_csv(here("data/options/ssp_anova_options_bayesian.csv")) |> 
-  mutate(iter = row_number())
+bayes_anova_options <- readr::read_csv(here("data/options/ssp_anova_options_bayesian.csv")) %>% 
+  mutate(row_id = row_number())
 
 # Set file directory -----------------------------------------------------------
 
@@ -43,7 +43,7 @@ n_batches <- 75
 
 # First, we split all the iteration as list
 bayes_anova_options_split <- 
-  split(bayes_anova_options, bayes_anova_options$iter)
+  split(bayes_anova_options, bayes_anova_options$row_id)
 
 # Since each iteration has 75 possible configuration, then we need:
 n_saves <- ceiling(length(bayes_anova_options_split) / n_batches)

data-raw/bf_anova_precalculations_results.R

@@ -1,100 +1,46 @@
 # Load necessary packages
 library(tidyverse)
 
-# Input params are not saved with result outputs thus we will join them here
-bayes_anova_options <-
-  tidyr::expand_grid(
-    m11 = 0,
-    m12 = seq(0, 1, by = 0.25),
-    m21 = seq(0, 1, by = 0.25),
-    m22 = seq(0, 1, by = 0.25),
-    tpr = seq(0.5, 0.95, by = 0.05),
-    effect = c("Main Effect 1", "Main Effect 2", "Interaction Effect"),
-    thresh = c(3, 6, 10),
-    prior_scale = c(1 / sqrt(2), 1, sqrt(2))
-  ) %>%
-  dplyr::filter(m12 <= m21) %>%
-  dplyr::slice(-1) %>%
-  dplyr::rowwise() %>%
-  dplyr::mutate(mu = mapply(c, m11, m12, m21, m22, SIMPLIFY = FALSE)) %>%
-  dplyr::ungroup() %>%
-  dplyr::mutate(iter = row_number()) |> 
-  dplyr::mutate(batch_id = str_pad(ceiling(row_number() / 75), width = 3, pad = "0"),
-               intra_batch_id = str_pad(((iter - 1) %% 75) + 1, width = 3, pad = "0"))
-
 # Read the outputs of all iterations
-bayes_anova_output <- tibble(filename = list.files(path = "./data/bayes-anova-res/", pattern = "\\.rds$", full.names = TRUE)) %>%
+bayes_anova_data <-
+  tibble(filename = list.files(
+    path = "./data/bayes-anova-res/",
+    pattern = "\\.rds$",
+    full.names = TRUE
+  )) %>%
   mutate(batch_id = str_extract(filename, "(?<=set-)[\\d]+") %>% str_pad(width = 3, pad = "0")) %>%
   mutate(data = map(filename, readRDS)) %>%
   unnest(data) %>%
   group_by(batch_id) %>%
-  mutate(intra_batch_id = str_pad(row_number(), width = 3, pad = "0")) %>%
+  mutate(calculation_id = str_pad(row_number(), width = 3, pad = "0")) %>%
   ungroup() %>%
-  mutate(result = map(data, "result"),
-         error = map(data, "error"),
-         error_message = map_chr(error, ~ pluck(.x, "message", .default = NA_character_)),
-         result_not_null = if_else(map_lgl(result, ~ !is.null(.x)),
-                                   1L,
-                                   0L),
-         n1 = map_dbl(result, ~ pluck(.x, "n1", .default = NA_real_)),
-         tpr_out = map_dbl(result, ~ pluck(.x, "tpr_out", .default = NA_real_)),
-         effect_out = map_chr(result, ~ pluck(.x, "effect", .default = NA_character_)),
-         # Because the iteration ids in the filenames do not have trailing zeros to keep the order to join the results with the input params we need to add them here
-         batch_id = str_extract(filename, "\\d+"),
-         batch_id = str_pad(batch_id, width = 3, pad = "0")
-         ) |> 
-  select(-data)
-
-bayes_anova_data <- left_join( bayes_anova_output, bayes_anova_options, by = c("batch_id", "intra_batch_id")) |> 
-  arrange(batch_id, intra_batch_id)
-
-# Randomly checking if some results match by recalculating them here
-source(here::here("R/ssp_bayesian_anova.R"))
-source(here::here("R/tpr_optim.R"))
-
-# Create a function to safely run Bayesian Anova
-safe_ssp_anova_bf <- purrr::safely(ssp_anova_bf)
-
-# For batch_id 001 and intra_batch_id 018
-safe_ssp_anova_bf(
-  tpr = 0.50,
-  effect = "Interaction Effect",
-  thresh = 3,
-  prior_scale = 0.7071068,
-  iter = 1000,
-  max_n = 500,
-  mu = c(0, 0, 0, 0),
-  sigma = 1
-)
-
-# For batch_id 010 and intra_batch_id 071
-safe_ssp_anova_bf(
-  tpr = 0.85,
-  effect = "Main Effect 2",
-  thresh = 10,
-  prior_scale = 1.4142136,
-  iter = 1000,
-  max_n = 500,
-  mu = c(0, 0, 0, 0.5),
-  sigma = 1
-)
-
-# For batch_id 015 and intra_batch_id 029
-safe_ssp_anova_bf(
-  tpr = 0.95,
-  effect = "Interaction Effect",
-  thresh = 10,
-  prior_scale = 1.4142136,
-  iter = 1000,
-  max_n = 500,
-  mu = c(0, 0, 0, 0.75),
-  sigma = 1
-)
-
-# Prepare data for saving
-# bayes_anova_precalculation_results <- 
-#   bayes_anova_data %>% 
-#   select(iterate, tpr_out, delta, thresh, prior_scale, n1, h0, ha, error_message) %>% 
-#   arrange(iterate)
+  mutate(parameters = map(data, "parameters"),
+         output = map(data, "output")) |>
+  mutate(
+    tpr_in = map_dbl(parameters, "tpr"),
+    effect = map_chr(parameters, "effect"),
+    thresh = map_int(parameters, "thresh"),
+    prior_scale = map_dbl(parameters, "prior_scale"),
+    mu = map(parameters, "mu"),
+    sigma = map_dbl(parameters, "sigma"),
+    result = map(output, "result"),
+    error = map(output, "error"),
+    error_message = map_chr(error, ~ pluck(.x, "message", .default = NA_character_)),
+    result_not_null = if_else(map_lgl(result, ~ !is.null(.x)),
+                              1L,
+                              0L),
+    n1 = map_dbl(result, ~ pluck(.x, "n1", .default = NA_real_)),
+    tpr_out = map_dbl(result, ~ pluck(.x, "tpr_out", .default = NA_real_)),
+    effect_out = map_chr(result, ~ pluck(.x, "effect", .default = NA_character_)),
+    # Because the iteration ids in the filenames do not have trailing zeros to keep the order to join the results with the input params we need to add them here
+    batch_id = str_extract(filename, "\\d+"),
+    batch_id = str_pad(batch_id, width = 3, pad = "0")
+  ) |>
+  select(-data, -parameters, -output, -result, -error) |> 
+  arrange(batch_id, calculation_id)
+
+# Calculating the number of calculated values
+bayes_anova_data |> 
+  count(result_not_null)
 
 usethis::use_data(bayes_anova_data, overwrite = TRUE)