added behavioral analysis scripts for error confidence

Author: Buch

code/analyses-behavioral/03_confidence-ratings.R

@@ -0,0 +1,315 @@
+# 03_confidence-ratings.r - error confidence rating analyses
+# author: [your name]
+
+# this script assumes it's sourced from batch_behavioral-analyses.r
+# with data & directories already loaded
+
+library(tidyverse)
+library(ez)
+library(effsize)
+
+cat("\n=== CONFIDENCE RATING ANALYSES ===\n")
+cat("started:", as.character(Sys.time()), "\n\n")
+
+# create output directory for confidence analyses
+confidence_dir <- file.path(session_dir, "confidence_ratings")
+dir.create(confidence_dir, showWarnings = FALSE)
+
+# === prepare confidence data ===
+prepare_confidence_data <- function(data, condition_type, verbose = TRUE) {
+  # prepare data for confidence analyses
+  #
+  # inputs:
+  #   data - trial-level behavioral data
+  #   condition_type - "visible_correctness", "visible_error_type", or "invisible_response_type"
+  #   verbose - print info
+  #
+  # outputs:
+  #   tibble ready for anova
+  
+  if (condition_type == "visible_correctness") {
+    # visible: correct vs error × social vs nonsocial
+    conf_data <- data %>%
+      filter(
+        code %in% c(111, 112, 113, 211, 212, 213),
+        !is.na(confidenceRating)
+      ) %>%
+      mutate(
+        subject = factor(subject),
+        social = factor(ifelse(code %in% c(111, 112, 113), "social", "nonsocial")),
+        correctness = factor(ifelse(code %in% c(111, 211), "correct", "error"))
+      ) %>%
+      select(subject, social, correctness, confidenceRating)
+    
+  } else if (condition_type == "visible_error_type") {
+    # visible: flanker vs nonflanker error × social vs nonsocial (errors only)
+    conf_data <- data %>%
+      filter(
+        code %in% c(112, 113, 212, 213),  # errors only
+        !is.na(confidenceRating)
+      ) %>%
+      mutate(
+        subject = factor(subject),
+        social = factor(ifelse(code %in% c(112, 113), "social", "nonsocial")),
+        error_type = factor(ifelse(code %in% c(112, 212), "flanker", "nonflanker"))
+      ) %>%
+      select(subject, social, error_type, confidenceRating)
+    
+  } else if (condition_type == "invisible_response_type") {
+    # invisible: flanker error vs nonflanker guess × social vs nonsocial
+    conf_data <- data %>%
+      filter(
+        code %in% c(102, 104, 202, 204),
+        !is.na(confidenceRating)
+      ) %>%
+      mutate(
+        subject = factor(subject),
+        social = factor(ifelse(code %in% c(102, 104), "social", "nonsocial")),
+        response_type = factor(ifelse(code %in% c(102, 202), "flanker_error", "nonflanker_guess"))
+      ) %>%
+      select(subject, social, response_type, confidenceRating)
+  }
+  
+  if (verbose) {
+    cat("  prepared", nrow(conf_data), "trials for", condition_type, "\n")
+  }
+  
+  return(conf_data)
+}
+
+# === run confidence analyses ===
+
+# 1. visible: correctness × social
+if (CONFIDENCE_ANALYSES$visible_correctness) {
+  cat("\n1. visible condition: correctness × social\n")
+  cat("   2×2 anova: correct/error × social/nonsocial\n")
+  
+  # prepare data
+  conf_correctness <- prepare_confidence_data(transformed_data, "visible_correctness")
+  
+  # calculate descriptives
+  descriptives_correctness <- conf_correctness %>%
+    group_by(social, correctness) %>%
+    summarise(
+      n = n(),
+      mean = mean(confidenceRating, na.rm = TRUE),
+      sd = sd(confidenceRating, na.rm = TRUE),
+      se = sd / sqrt(n),
+      .groups = "drop"
+    )
+  
+  cat("\n   descriptive statistics:\n")
+  print(descriptives_correctness)
+  
+  # run 2×2 repeated measures anova
+  anova_correctness <- ezANOVA(
+    data = conf_correctness,
+    dv = confidenceRating,
+    wid = subject,
+    within = .(social, correctness),
+    detailed = TRUE,
+    type = 3
+  )
+  
+  cat("\n   anova results:\n")
+  print(anova_correctness$ANOVA)
+  
+  # extract ANOVA table & calculate partial eta squared
+  anova_table_correctness <- anova_correctness$ANOVA %>%
+    mutate(
+      partial_eta_sq = SSn / (SSn + SSd)
+    )
+  
+  # save results
+  results_correctness <- list(
+    descriptives = descriptives_correctness,
+    anova = anova_correctness,
+    anova_table = anova_table_correctness,
+    data = conf_correctness
+  )
+  saveRDS(results_correctness, file.path(confidence_dir, "visible_correctness.rds"))
+  
+  # save tables to tables directory
+  save_descriptives_table(descriptives_correctness, "confidence_visible_correctness_descriptives", tables_dir)
+  save_anova_results(anova_table_correctness, "confidence_visible_correctness", tables_dir)
+  
+  # create plot
+  plot_path <- file.path(confidence_dir, "confidence_visible_correctness.png")
+  plot_confidence_interaction(conf_correctness %>%
+                                mutate(is_error = correctness == "error"),
+                              plot_type = "correctness",
+                              save_path = plot_path)
+}
+
+# 2. visible: error type × social (errors only)
+if (CONFIDENCE_ANALYSES$visible_error_type) {
+  cat("\n2. visible condition: error type × social (errors only)\n")
+  cat("   2×2 anova: flanker/nonflanker × social/nonsocial\n")
+  
+  # prepare data
+  conf_error_type <- prepare_confidence_data(transformed_data, "visible_error_type")
+  
+  # check sample sizes per cell
+  cell_counts <- conf_error_type %>%
+    group_by(subject, social, error_type) %>%
+    summarise(n_trials = n(), .groups = "drop") %>%
+    group_by(social, error_type) %>%
+    summarise(n_subjects = n(), .groups = "drop")
+  
+  cat("\n   subjects per cell:\n")
+  print(cell_counts)
+  
+  # only run if sufficient data
+  if (all(cell_counts$n_subjects >= 10)) {
+    # calculate descriptives
+    descriptives_error_type <- conf_error_type %>%
+      group_by(social, error_type) %>%
+      summarise(
+        n = n(),
+        mean = mean(confidenceRating, na.rm = TRUE),
+        sd = sd(confidenceRating, na.rm = TRUE),
+        se = sd / sqrt(n),
+        .groups = "drop"
+      )
+    
+    cat("\n   descriptive statistics:\n")
+    print(descriptives_error_type)
+    
+    # run 2×2 repeated measures anova
+    anova_error_type <- ezANOVA(
+      data = conf_error_type,
+      dv = confidenceRating,
+      wid = subject,
+      within = .(social, error_type),
+      detailed = TRUE,
+      type = 3
+    )
+    
+    cat("\n   anova results:\n")
+    print(anova_error_type$ANOVA)
+    
+    # extract ANOVA table & calculate partial eta squared
+    anova_table_error_type <- anova_error_type$ANOVA %>%
+      mutate(
+        partial_eta_sq = SSn / (SSn + SSd)
+      )
+    
+    # save results
+    results_error_type <- list(
+      descriptives = descriptives_error_type,
+      anova = anova_error_type,
+      anova_table = anova_table_error_type,
+      data = conf_error_type
+    )
+    saveRDS(results_error_type, file.path(confidence_dir, "visible_error_type.rds"))
+    
+    # save tables to tables directory
+    save_descriptives_table(descriptives_error_type, "confidence_visible_error_type_descriptives", tables_dir)
+    save_anova_results(anova_table_error_type, "confidence_visible_error_type", tables_dir)
+    
+    # create plot
+    plot_path <- file.path(confidence_dir, "confidence_visible_error_type.png")
+    plot_confidence_interaction(conf_error_type %>%
+                                  mutate(is_flanker = error_type == "flanker"),
+                                plot_type = "error_type",
+                                save_path = plot_path)
+    
+  } else {
+    cat("\n   insufficient data for error type analysis\n")
+  }
+}
+
+# 3. invisible: response type × social
+if (CONFIDENCE_ANALYSES$invisible_response_type) {
+  cat("\n3. invisible condition: response type × social\n")
+  cat("   2×2 anova: flanker error/nonflanker guess × social/nonsocial\n")
+  
+  # prepare data
+  conf_response_type <- prepare_confidence_data(transformed_data, "invisible_response_type")
+  
+  # calculate descriptives
+  descriptives_response_type <- conf_response_type %>%
+    group_by(social, response_type) %>%
+    summarise(
+      n = n(),
+      mean = mean(confidenceRating, na.rm = TRUE),
+      sd = sd(confidenceRating, na.rm = TRUE),
+      se = sd / sqrt(n),
+      .groups = "drop"
+    )
+  
+  cat("\n   descriptive statistics:\n")
+  print(descriptives_response_type)
+  
+  # run 2×2 repeated measures anova
+  anova_response_type <- ezANOVA(
+    data = conf_response_type,
+    dv = confidenceRating,
+    wid = subject,
+    within = .(social, response_type),
+    detailed = TRUE,
+    type = 3
+  )
+  
+  cat("\n   anova results:\n")
+  print(anova_response_type$ANOVA)
+  
+  # extract ANOVA table & calculate partial eta squared
+  anova_table_response_type <- anova_response_type$ANOVA %>%
+    mutate(
+      partial_eta_sq = SSn / (SSn + SSd)
+    )
+  
+  # save results
+  results_response_type <- list(
+    descriptives = descriptives_response_type,
+    anova = anova_response_type,
+    anova_table = anova_table_response_type,
+    data = conf_response_type
+  )
+  saveRDS(results_response_type, file.path(confidence_dir, "invisible_response_type.rds"))
+  
+  # save tables to tables directory
+  save_descriptives_table(descriptives_response_type, "confidence_invisible_response_type_descriptives", tables_dir)
+  save_anova_results(anova_table_response_type, "confidence_invisible_response_type", tables_dir)
+  
+  # create plot
+  plot_path <- file.path(confidence_dir, "confidence_invisible_response_type.png")
+  plot_confidence_interaction(conf_response_type %>%
+                                mutate(is_flanker_error = response_type == "flanker_error"),
+                              plot_type = "response_type",
+                              save_path = plot_path)
+}
+
+# === save summary table ===
+cat("\n=== saving confidence rating summary tables ===\n")
+
+# combine all descriptives for summary
+all_descriptives <- list()
+
+if (exists("descriptives_correctness")) {
+  all_descriptives$visible_correctness <- descriptives_correctness %>%
+    mutate(analysis = "Visible: Correctness")
+}
+
+if (exists("descriptives_error_type")) {
+  all_descriptives$visible_error_type <- descriptives_error_type %>%
+    mutate(analysis = "Visible: Error Type")
+}
+
+if (exists("descriptives_response_type")) {
+  all_descriptives$invisible_response_type <- descriptives_response_type %>%
+    mutate(analysis = "Invisible: Response Type")
+}
+
+if (length(all_descriptives) > 0) {
+  summary_table <- bind_rows(all_descriptives) %>%
+    mutate(across(c(mean, sd, se), ~round(., 3)))
+  
+  # save as csv
+  write_csv(summary_table, file.path(tables_dir, "confidence_summary.csv"))
+  cat("   summary table saved to confidence_summary.csv\n")
+}
+
+cat("\n=== confidence rating analyses complete ===\n")
+cat("results saved to:", confidence_dir, "\n")