---
title: "Effect size computations: shurakovndstakeseffectnew"
format:
  html:
    toc: true
execute:
  echo: true
  warning: true
  message: false
---

Computes standardized mean differences (`d`) and sampling variances (`v`) for the
extraction YAML `papers/shurakovndstakeseffectnew/shurakovndstakeseffectnew.yaml`.

Also reproduces the Table 2 composite-score descriptives and Cohen's *d* from the
OSF data files in `papers/shurakovndstakeseffectnew/data/`, plus a full-valid
sample sensitivity analysis that keeps all valid responses surviving the paper's
stated exclusion rules.

## Inputs (from extraction)

```{r}
paper_key <- "shurakovndstakeseffectnew"
sign_convention <- "d = mean(low) - mean(high)"

effects <- list(
  list(
    study_id = 1,
    effect_id = "s1_e1",
    method_used = "between_reported_d_n",
    n_low = 100,
    n_high = 100,
    mean_low = 5.32,
    mean_high = 0.28,
    reported_d = 1.06,
    notes_on_assumptions = "Experiment 1 (first-person): d from Table 2 (Tukey HSD); n=100 per condition (first 100 valid responses per condition)."
  ),
  list(
    study_id = 2,
    effect_id = "s2_e1",
    method_used = "between_reported_d_n",
    n_low = 100,
    n_high = 100,
    mean_low = 4.64,
    mean_high = -0.94,
    reported_d = 1.28,
    notes_on_assumptions = "Experiment 1 (third-person): d from Table 2 (Tukey HSD); n=100 per condition (first 100 valid responses per condition)."
  ),
  list(
    study_id = 3,
    effect_id = "s3_e1",
    method_used = "between_reported_d_n",
    n_low = 100,
    n_high = 100,
    mean_low = 5.36,
    mean_high = 1.59,
    reported_d = 0.94,
    notes_on_assumptions = "Experiment 2 (modified design; first-person): d from Table 2 (Tukey HSD); n=100 per condition (first 100 valid responses per condition)."
  )
)
```

## Computation (d and v)

```{r}
default_or <- function(x, default) {
  if (is.null(x)) default else x
}

stop_if_missing <- function(x, name) {
  if (is.na(x)) stop(sprintf("Missing required input: %s", name), call. = FALSE)
}

var_d_independent <- function(d, n_high, n_low) {
  n <- n_high + n_low
  (n / (n_high * n_low)) + (d^2 / (2 * (n - 2)))
}

infer_sign <- function(mean_low, mean_high) {
  if (!is.na(mean_low) && !is.na(mean_high) && mean_low != mean_high) {
    return(sign(mean_low - mean_high))
  }
  stop("Cannot infer sign from means", call. = FALSE)
}

compute_effect <- function(effect_inputs) {
  study_id <- effect_inputs$study_id
  effect_id <- effect_inputs$effect_id
  method_used <- effect_inputs$method_used

  n_high <- default_or(effect_inputs$n_high, NA_integer_)
  n_low <- default_or(effect_inputs$n_low, NA_integer_)
  mean_high <- default_or(effect_inputs$mean_high, NA_real_)
  mean_low <- default_or(effect_inputs$mean_low, NA_real_)
  reported_d <- default_or(effect_inputs$reported_d, NA_real_)
  notes_on_assumptions <- default_or(effect_inputs$notes_on_assumptions, "")

  if (method_used != "between_reported_d_n") {
    stop(sprintf("Unknown method_used: %s", method_used), call. = FALSE)
  }

  stop_if_missing(reported_d, "reported_d")
  stop_if_missing(n_high, "n_high")
  stop_if_missing(n_low, "n_low")

  sign_used <- infer_sign(mean_low, mean_high)
  d <- sign_used * abs(reported_d)
  v <- var_d_independent(d = d, n_high = n_high, n_low = n_low)

  inputs_used <- paste(
    c(
      sprintf("method=%s", method_used),
      sprintf("sign_convention=%s", sign_convention),
      sprintf("n_low=%s", n_low),
      sprintf("n_high=%s", n_high),
      sprintf("mean_low=%s", mean_low),
      sprintf("mean_high=%s", mean_high),
      sprintf("reported_d=%s", reported_d)
    ),
    collapse = ", "
  )

  data.frame(
    paper_key = paper_key,
    study_id = study_id,
    effect_id = effect_id,
    design = "Between-Subjects",
    method_used = method_used,
    inputs_used = inputs_used,
    d = d,
    v = v,
    notes_on_assumptions = notes_on_assumptions,
    stringsAsFactors = FALSE
  )
}

audit <- do.call(rbind, lapply(effects, compute_effect))
audit
```

## Paste-ready YAML snippets

```{r}
for (i in seq_len(nrow(audit))) {
  row <- audit[i, ]
  cat(sprintf("\n# %s (study_id=%s)\n", row$effect_id, row$study_id))
  cat(sprintf(
    "effect_size:\n  metric: SMD\n  d: %.12f\n  v: %.12f\n  computed_from: reported_d\n  needs_review: false\n  notes: \"%s\"\n",
    row$d,
    row$v,
    gsub("\"", "'", row$inputs_used)
  ))
}
```

## Reproduce Table 2 (OSF data check)

```{r}
read_csv_local <- function(path) {
  read.csv(path, na.strings = c("", " "), check.names = FALSE)
}

composite_from_cols <- function(df, col_a, col_b, n_take = NULL) {
  a <- suppressWarnings(as.numeric(df[[col_a]]))
  b <- suppressWarnings(as.numeric(df[[col_b]]))
  keep <- !is.na(a) & !is.na(b)
  a <- a[keep]
  b <- b[keep]

  if (!is.null(n_take)) {
    a <- head(a, n_take)
    b <- head(b, n_take)
  }

  # In OSF data, A is coded 1="I do", 2="I don't". Convert to +1 / -1.
  sign <- ifelse(a == 1, 1, ifelse(a == 2, -1, NA_real_))
  if (any(is.na(sign))) stop(sprintf("Unexpected A values in %s", col_a), call. = FALSE)
  sign * b
}

summarize_conditions <- function(df, mapping, n_take = NULL) {
  composites <- list()
  for (nm in names(mapping)) {
    cols <- mapping[[nm]]
    composites[[nm]] <- composite_from_cols(df, cols$a, cols$b, n_take = n_take)
  }
  summarize_composites(composites)
}

summarize_composites <- function(composites) {
  rows <- list()
  for (nm in names(composites)) {
    comp <- composites[[nm]]
    rows[[nm]] <- data.frame(
      condition = nm,
      n = length(comp),
      mean = mean(comp),
      sd = sd(comp),
      stringsAsFactors = FALSE
    )
  }
  do.call(rbind, rows)
}

exp1 <- read_csv_local("../data/Experiment_1_analized_responses.csv")
exp2_first100 <- read_csv_local("../data/Experiment_2_analyzed_responses.csv")
exp2_all <- read_csv_local("../data/Experiment_2_all_responses.csv")

exp1_first_map <- list(
  Neutral = list(a = "BA_1_N_A", b = "BA_1_N_B"),
  Stakes = list(a = "BA_1_S_A", b = "BA_1_S_B"),
  Evidence = list(a = "BA_1_E_A", b = "BA_1_E_B")
)
exp1_third_map <- list(
  Neutral = list(a = "BA_3_N_A", b = "BA_3_N_B"),
  Stakes = list(a = "BA_3_S_A", b = "BA_3_S_B"),
  Evidence = list(a = "BA_3_E_A", b = "BA_3_E_B")
)
exp2_mod_map <- list(
  Neutral = list(a = "BA_1_N_A", b = "BA_1_N_B"),
  Stakes = list(a = "BA_1_S_A", b = "BA_1_S_B"),
  Evidence = list(a = "BA_1_E_A", b = "BA_1_E_B")
)

target_n <- 100

stats_e1_first <- summarize_conditions(exp1, exp1_first_map, n_take = target_n)
stats_e1_third <- summarize_conditions(exp1, exp1_third_map, n_take = target_n)
stats_e2_mod <- summarize_conditions(exp2_first100, exp2_mod_map, n_take = target_n)

pooled_sd <- function(n_high, n_low, sd_high, sd_low) {
  sqrt(((n_high - 1) * sd_high^2 + (n_low - 1) * sd_low^2) / (n_high + n_low - 2))
}

cohen_d_from_group_stats <- function(stats, low = "Neutral", high = "Stakes") {
  low_row <- stats[stats$condition == low, ]
  high_row <- stats[stats$condition == high, ]
  if (nrow(low_row) != 1 || nrow(high_row) != 1) stop("Missing condition rows", call. = FALSE)
  (low_row$mean - high_row$mean) / pooled_sd(
    n_high = high_row$n,
    n_low = low_row$n,
    sd_high = high_row$sd,
    sd_low = low_row$sd
  )
}

d_e1_first <- cohen_d_from_group_stats(stats_e1_first)
d_e1_third <- cohen_d_from_group_stats(stats_e1_third)
d_e2_mod <- cohen_d_from_group_stats(stats_e2_mod)

reported <- data.frame(
  subset = c("Experiment 1 (first-person)", "Experiment 1 (third-person)", "Experiment 2 (modified design)"),
  mean_neutral = c(5.32, 4.64, 5.36),
  sd_neutral = c(2.93, 3.53, 2.63),
  mean_stakes = c(0.28, -0.94, 1.59),
  sd_stakes = c(5.32, 5.04, 5.06),
  d_stakes_vs_neutral = c(1.06, 1.28, 0.94),
  stringsAsFactors = FALSE
)

computed <- data.frame(
  subset = c("Experiment 1 (first-person)", "Experiment 1 (third-person)", "Experiment 2 (modified design)"),
  mean_neutral = c(stats_e1_first$mean[stats_e1_first$condition == "Neutral"], stats_e1_third$mean[stats_e1_third$condition == "Neutral"], stats_e2_mod$mean[stats_e2_mod$condition == "Neutral"]),
  sd_neutral = c(stats_e1_first$sd[stats_e1_first$condition == "Neutral"], stats_e1_third$sd[stats_e1_third$condition == "Neutral"], stats_e2_mod$sd[stats_e2_mod$condition == "Neutral"]),
  mean_stakes = c(stats_e1_first$mean[stats_e1_first$condition == "Stakes"], stats_e1_third$mean[stats_e1_third$condition == "Stakes"], stats_e2_mod$mean[stats_e2_mod$condition == "Stakes"]),
  sd_stakes = c(stats_e1_first$sd[stats_e1_first$condition == "Stakes"], stats_e1_third$sd[stats_e1_third$condition == "Stakes"], stats_e2_mod$sd[stats_e2_mod$condition == "Stakes"]),
  d_stakes_vs_neutral = c(d_e1_first, d_e1_third, d_e2_mod),
  stringsAsFactors = FALSE
)

merge(reported, computed, by = "subset", suffixes = c("_reported", "_from_data"))
```

## Full-valid sensitivity analysis (OSF data)

The paper's confirmatory analyses used the first 100 valid responses per condition.
For meta-analytic sensitivity analysis, we can instead retain all valid responses
surviving the paper's stated exclusion rules:

- Experiment 1: all rows in `Experiment_1_analized_responses.csv` (already filtered
  for failed attention checks).
- Experiment 2: rows in `Experiment_2_all_responses.csv` that both
  endorsed the knowledge ascription (`Set-up = 1`) and passed the
  condition-specific attention check (`*_AT = 2`).

```{r}
subset_exp2_full_valid <- function(df, prefix) {
  keep <- df[["Set-up"]] == 1 & df[[paste0(prefix, "_AT")]] == 2
  df[keep, , drop = FALSE]
}

full_valid_exp2_composites <- list(
  Neutral = composite_from_cols(
    subset_exp2_full_valid(exp2_all, "BA_1_N"),
    "BA_1_N_A",
    "BA_1_N_B"
  ),
  Stakes = composite_from_cols(
    subset_exp2_full_valid(exp2_all, "BA_1_S"),
    "BA_1_S_A",
    "BA_1_S_B"
  ),
  Evidence = composite_from_cols(
    subset_exp2_full_valid(exp2_all, "BA_1_E"),
    "BA_1_E_A",
    "BA_1_E_B"
  )
)

stats_e1_first_full <- summarize_conditions(exp1, exp1_first_map)
stats_e1_third_full <- summarize_conditions(exp1, exp1_third_map)
stats_e2_mod_full <- summarize_composites(full_valid_exp2_composites)

full_valid_stats <- rbind(
  cbind(subset = "Experiment 1 (first-person, full valid sample sensitivity)", stats_e1_first_full),
  cbind(subset = "Experiment 1 (third-person, full valid sample sensitivity)", stats_e1_third_full),
  cbind(subset = "Experiment 2 (modified design, full valid sample sensitivity)", stats_e2_mod_full)
)

full_valid_stats
```

## First-100 vs full-valid comparison

```{r}
build_effect_from_stats <- function(study_id, effect_id, stats, notes_on_assumptions) {
  low_row <- stats[stats$condition == "Neutral", ]
  high_row <- stats[stats$condition == "Stakes", ]
  if (nrow(low_row) != 1 || nrow(high_row) != 1) stop("Missing condition rows", call. = FALSE)

  d <- cohen_d_from_group_stats(stats)
  v <- var_d_independent(d = d, n_high = high_row$n, n_low = low_row$n)

  inputs_used <- paste(
    c(
      "method=between_group_stats",
      sprintf("sign_convention=%s", sign_convention),
      sprintf("n_low=%s", low_row$n),
      sprintf("n_high=%s", high_row$n),
      sprintf("mean_low=%s", low_row$mean),
      sprintf("mean_high=%s", high_row$mean),
      sprintf("sd_low=%s", low_row$sd),
      sprintf("sd_high=%s", high_row$sd)
    ),
    collapse = ", "
  )

  data.frame(
    paper_key = paper_key,
    study_id = study_id,
    effect_id = effect_id,
    design = "Between-Subjects",
    method_used = "between_group_stats",
    inputs_used = inputs_used,
    d = d,
    v = v,
    notes_on_assumptions = notes_on_assumptions,
    stringsAsFactors = FALSE
  )
}

audit_full_valid <- do.call(
  rbind,
  list(
    build_effect_from_stats(
      study_id = 4,
      effect_id = "s4_e1",
      stats = stats_e1_first_full,
      notes_on_assumptions = "Experiment 1 (first-person, full valid sample sensitivity): all valid OSF analyzed responses after the paper's attention-check exclusion."
    ),
    build_effect_from_stats(
      study_id = 5,
      effect_id = "s5_e1",
      stats = stats_e1_third_full,
      notes_on_assumptions = "Experiment 1 (third-person, full valid sample sensitivity): all valid OSF analyzed responses after the paper's attention-check exclusion."
    ),
    build_effect_from_stats(
      study_id = 6,
      effect_id = "s6_e1",
      stats = stats_e2_mod_full,
      notes_on_assumptions = "Experiment 2 (modified design, full valid sample sensitivity): all valid OSF raw responses after applying the paper's stated exclusion rules (knowledge endorsers only; condition-specific attention-check pass)."
    )
  )
)

audit_first100_data <- do.call(
  rbind,
  list(
    build_effect_from_stats(
      study_id = 1,
      effect_id = "s1_e1_first100_data",
      stats = stats_e1_first,
      notes_on_assumptions = "Experiment 1 (first-person): first 100 valid OSF analyzed responses per condition."
    ),
    build_effect_from_stats(
      study_id = 2,
      effect_id = "s2_e1_first100_data",
      stats = stats_e1_third,
      notes_on_assumptions = "Experiment 1 (third-person): first 100 valid OSF analyzed responses per condition."
    ),
    build_effect_from_stats(
      study_id = 3,
      effect_id = "s3_e1_first100_data",
      stats = stats_e2_mod,
      notes_on_assumptions = "Experiment 2 (modified design): first 100 valid OSF responses per condition."
    )
  )
)

comparison <- data.frame(
  subset = c("Experiment 1 (first-person)", "Experiment 1 (third-person)", "Experiment 2 (modified design)"),
  n_low_first100 = c(
    stats_e1_first$n[stats_e1_first$condition == "Neutral"],
    stats_e1_third$n[stats_e1_third$condition == "Neutral"],
    stats_e2_mod$n[stats_e2_mod$condition == "Neutral"]
  ),
  n_high_first100 = c(
    stats_e1_first$n[stats_e1_first$condition == "Stakes"],
    stats_e1_third$n[stats_e1_third$condition == "Stakes"],
    stats_e2_mod$n[stats_e2_mod$condition == "Stakes"]
  ),
  d_first100 = audit_first100_data$d,
  v_first100 = audit_first100_data$v,
  n_low_full_valid = c(
    stats_e1_first_full$n[stats_e1_first_full$condition == "Neutral"],
    stats_e1_third_full$n[stats_e1_third_full$condition == "Neutral"],
    stats_e2_mod_full$n[stats_e2_mod_full$condition == "Neutral"]
  ),
  n_high_full_valid = c(
    stats_e1_first_full$n[stats_e1_first_full$condition == "Stakes"],
    stats_e1_third_full$n[stats_e1_third_full$condition == "Stakes"],
    stats_e2_mod_full$n[stats_e2_mod_full$condition == "Stakes"]
  ),
  d_full_valid = audit_full_valid$d,
  v_full_valid = audit_full_valid$v,
  stringsAsFactors = FALSE
)

comparison
```

## Paste-ready YAML snippets (full-valid sensitivity studies)

```{r}
for (i in seq_len(nrow(audit_full_valid))) {
  row <- audit_full_valid[i, ]
  cat(sprintf("\n# %s (study_id=%s)\n", row$effect_id, row$study_id))
  cat(sprintf(
    "effect_size:\n  metric: SMD\n  d: %.12f\n  v: %.12f\n  computed_from: groups\n  needs_review: false\n  notes: \"%s\"\n",
    row$d,
    row$v,
    gsub("\"", "'", row$inputs_used)
  ))
}
```