Add correlation wth RAPM without resampling reasoning tasks

Signed-off-by: Liang Zhang <[email protected]>

R/constants.R

@@ -1,6 +1,21 @@
 game_id_rapm <- bit64::as.integer64(265520726213317) # 瑞文高级推理
 # data for this game should be removed because another version was used
 game_id_rata <- bit64::as.integer64(240052750955077) # 远距离联想A (obsolete)
+game_id_reasoning <- bit64::as.integer64(
+  c(
+    355740871500677,
+    355740333933445,
+    356101783560965,
+    411281158706373,
+    324179001000709,
+    268008982659144,
+    324179964613381,
+    361750352892805,
+    324177880433541
+  )
+)
+# we define here to be used in g factor resampling
+num_vars_total <- 76
 
 # used in data quality check
 thresh_prop_miss <- 0.25

R/tar_prepare.R

@@ -1,10 +1,11 @@
-prepare_config_vars <- function(num_vars_total, n_steps) {
-  num_vars_base <- num_vars_total %/% n_steps
-  tibble::tibble(
-    num_vars = seq(num_vars_base, num_vars_total, num_vars_base),
-    use_pairs = num_vars * 2 <= num_vars_total
-  ) |>
+prepare_config_vars <- function(num_vars_total, ...,
+                                step = 3, from = 3, use_pairs = TRUE) {
+  out <- tibble::tibble(num_vars = seq(from, num_vars_total, step)) |>
     dplyr::filter(choose(num_vars_total, num_vars) > 1e4)
+  if (use_pairs) {
+    out$use_pairs <- out$num_vars * 2 <= num_vars_total
+  }
+  out
 }
 
 prepare_config_cpm <- function(...) {

_scripts/g_factor.R

@@ -7,10 +7,7 @@ tar_option_set(
 )
 setup_parallel_plan()
 
-n_vars_total <- 76
-n_steps <- 20
-
-config_vars <- prepare_config_vars(n_vars_total, n_steps)
+config_vars <- prepare_config_vars(num_vars_total)
 config_cpm <- prepare_config_cpm(
   config == "gsr",
   task == "wm",

_scripts/g_factor_cor_rapm.R

@@ -0,0 +1,90 @@
+library(targets)
+tar_source()
+tar_option_set(
+  packages = c("tidyverse", "bit64", "lavaan"),
+  format = "qs",
+  controller = setup_crew_controller("gf")
+)
+setup_parallel_plan()
+
+config_vars <- prepare_config_vars(
+  num_vars_total - length(game_id_reasoning),
+  use_pairs = FALSE
+)
+config_cpm <- prepare_config_cpm(
+  config == "gsr",
+  task == "wm",
+  run == "full",
+  thresh_method == "alpha",
+  thresh_level == 0.01
+)
+branches_g <- tarchetypes::tar_map(
+  config_vars,
+  tarchetypes::tar_rep(
+    vars_sample,
+    resample_vars(names(indices_cogstruct), num_vars),
+    batches = 10,
+    reps = 10,
+    iteration = "list",
+    deployment = "main"
+  ),
+  tarchetypes::tar_rep2(
+    fit_g,
+    lapply_tar_batches(
+      vars_sample,
+      fit_efa_g,
+      data = indices_cogstruct,
+      missing = "ml"
+    ),
+    vars_sample,
+    iteration = "list"
+  ),
+  tarchetypes::tar_rep2(
+    scores_g,
+    lapply_tar_batches(
+      fit_g,
+      extract_g_scores,
+      data = indices_cogstruct
+    ),
+    fit_g,
+    iteration = "list"
+  ),
+  tarchetypes::tar_rep2(
+    cor_rapm,
+    lapply_tar_batches(
+      scores_g,
+      \(x) {
+        indices_rapm |>
+          merge(x, by = "row.names") |>
+          summarise(r = cor(score, f1, use = "pairwise"))
+      }
+    ) |>
+      list_rbind_tar_batches(),
+    scores_g,
+    iteration = "list"
+  )
+)
+
+list(
+  tarchetypes::tar_file_read(
+    indices_cogstruct,
+    path_obj_from_proj("indices_cogstruct", "prepare_source_data"),
+    read = qs::qread(!!.x)
+  ),
+  tarchetypes::tar_file_read(
+    indices_rapm,
+    path_obj_from_proj("indices_rapm", "prepare_source_data"),
+    read = qs::qread(!!.x)
+  ),
+  branches_g,
+  tarchetypes::tar_combine(
+    cor_rapm,
+    branches_g$cor_rapm,
+    command = list(!!!.x) |>
+      lapply(bind_rows) |>
+      bind_rows_meta(
+        .names = names(config_vars),
+        .prefix = "cor_rapm"
+      )
+  )
+)

_targets.yaml

@@ -10,6 +10,9 @@ prepare_neural:
 g_factor:
   script: _scripts/g_factor.R
   store: _targets/g_factor
+g_factor_cor_rapm:
+  script: _scripts/g_factor_cor_rapm.R
+  store: _targets/g_factor_cor_rapm
 explore_factors:
   script: _scripts/explore_factors.R
   store: _targets/explore_factors