############################################
##  Multigroup Difference Analysis (ANOVA/Kruskal-Wallis)
############################################
#' @export
mda <- function(dataset,
                var,
                group,
                normality_type = c("overall", "by_group"),
                data_filter = "",
                envir = parent.frame()) {
  
  # 1. 基础参数处理
  normality_type <- match.arg(normality_type, choices = c("overall", "by_group"))
  df_name <- if (is_string(dataset)) dataset else deparse(substitute(dataset))
  
  # 2. 数据提取：只保留“因变量+分组变量”
  dataset <- get_data(
    dataset,
    vars = c(var, group),  # 强制只取2个核心变量，剔除冗余列
    filt = data_filter,
    na.rm = FALSE,         # 先不删缺失值，后续统一过滤
    envir = envir
  )
  
  # 3. 数据校验
  if (!var %in% colnames(dataset)) {
    stop(paste("因变量", var, "未在数据集中找到！"), call. = FALSE)
  }
  if (!group %in% colnames(dataset)) {
    stop(paste("分组变量", group, "未在数据集中找到！"), call. = FALSE)
  }
  if (!is.numeric(dataset[[var]])) {
    stop(paste("因变量", var, "必须是数值型（当前类型：", class(dataset[[var]]), "）！"), call. = FALSE)
  }
  
  # 4. 有效样本过滤：剔除任一变量缺失的样本
  valid_indices <- !is.na(dataset[[var]]) & !is.na(dataset[[group]])
  valid_data <- dataset[valid_indices, ]  # 仅保留有效样本的2列数据
  if (nrow(valid_data) == 0) {
    stop("无有效样本（所有样本的因变量/分组变量存在缺失值）！", call. = FALSE)
  }
  
  # 5. 分组变量处理：强制转因子+校验水平
  valid_data[[group]] <- as.factor(valid_data[[group]])  # 强制转因子，避免字符型干扰
  valid_levels <- length(levels(valid_data[[group]]))    # 用levels()确保因子水平正确
  if (valid_levels < 2) {
    stop(paste("分组变量有效水平不足2个（当前水平数：", valid_levels, "），无法执行检验！"), call. = FALSE)
  }
  
  # 6. 检验计算：调用辅助函数
  homo_res <- run_homo_test(valid_data, var, group)  # 方差齐性检验
  norm_res <- run_norm_test(valid_data, var, group, normality_type)  # 正态性检验
  
  # 7. 绘图数据准备
  plot_obj <- list(
    norm = list(
      data = valid_data[[var]],
      group_data = if (normality_type == "by_group") {
        # 把命名向量转为无命名列表，避免asJSON警告
        lapply(split(valid_data[[var]], valid_data[[group]]), function(x) x)
      } else NULL,
      var = var,
      group = group,
      type = normality_type
    ),
    homo = list(
      data = valid_data,
      var = var,
      group = group
    )
  )
  
  # 8. 结果打包：对齐单独检验的输出结构
  out <- structure(
    list(
      df_name = df_name,
      var = var,
      group = group,
      normality_type = normality_type,
      data_filter = if (data_filter == "") "None" else data_filter,
      valid_n = nrow(valid_data),  # 有效样本量
      homo_res = homo_res,         # 方差齐性检验结果
      norm_res = norm_res,         # 正态性检验结果
      plot_obj = plot_obj
    ),
    class = "mda"  
  )
  out
}

# ------------------------------
# 辅助函数1：方差齐性检验
# ------------------------------
run_homo_test <- function(valid_data, var, group) {
  x <- valid_data[[var]]
  g <- valid_data[[group]]
  res <- tibble::tibble(Test = character(), Statistic = numeric(), p.value = numeric())
  
  # 1. Levene检验
  if (requireNamespace("car", quietly = TRUE)) {
    tmp <- tryCatch(
      expr = car::leveneTest(x ~ g),
      error = function(e) {
        message(paste("Levene检验执行失败：", e$message))
        return(NULL)
      }
    )
    if (!is.null(tmp) && nrow(tmp) > 0) {
      res <- tibble::add_row(res,
                             Test = "Levene",
                             Statistic = as.numeric(tmp[["F value"]][1]),
                             p.value = as.numeric(tmp[["Pr(>F)"]][1]))
    }
  } else {
    res <- tibble::add_row(res,
                           Test = "Levene",
                           Statistic = NA_real_,
                           p.value = NA_real_)
    message("提示：需安装car包以运行Levene检验")
  }
  
  # 2. Bartlett检验
  tmp <- tryCatch(
    expr = stats::bartlett.test(x, g),
    error = function(e) {
      message(paste("Bartlett检验执行失败：", e$message))
      return(NULL)
    }
  )
  if (!is.null(tmp)) {
    res <- tibble::add_row(res,
                           Test = "Bartlett",
                           Statistic = as.numeric(tmp$statistic),
                           p.value = as.numeric(tmp$p.value))
  } else {
    res <- tibble::add_row(res,
                           Test = "Bartlett",
                           Statistic = NA_real_,
                           p.value = NA_real_)
  }
  
  # 3. Fligner检验
  tmp <- tryCatch(
    expr = stats::fligner.test(x, g),
    error = function(e) {
      message(paste("Fligner检验执行失败：", e$message))
      return(NULL)
    }
  )
  if (!is.null(tmp)) {
    res <- tibble::add_row(res,
                           Test = "Fligner",
                           Statistic = as.numeric(tmp$statistic),
                           p.value = as.numeric(tmp$p.value))
  } else {
    res <- tibble::add_row(res,
                           Test = "Fligner",
                           Statistic = NA_real_,
                           p.value = NA_real_)
  }
  
  res
}

# ------------------------------
# 辅助函数2：正态性检验
# ------------------------------
run_norm_test <- function(valid_data, var, group, normality_type) {
  x <- valid_data[[var]]
  g <- valid_data[[group]]
  res <- tibble::tibble(Group = character(), Test = character(), Statistic = numeric(), p.value = numeric())
  
  # 1. 整体正态性检验
  if (normality_type == "overall") {
    res <- dplyr::bind_rows(res, get_single_norm(x, group_label = "Overall"))
  }
  
  # 2. 按分组正态性检验
  if (normality_type == "by_group") {
    for (level in levels(g)) {
      group_x <- x[g == level]
      res <- dplyr::bind_rows(res, get_single_norm(group_x, group_label = level))
    }
  }
  
  res
}

# ------------------------------
# 辅助函数3：单组正态性检验
# ------------------------------
get_single_norm <- function(x, group_label) {
  res <- tibble::tibble(Group = group_label, Test = character(), Statistic = numeric(), p.value = numeric())
  n <- length(x)
  
  # 1. Shapiro-Wilk检验
  if (n >= 3 && n <= 5000) {
    tmp <- tryCatch(
      expr = stats::shapiro.test(x),
      error = function(e) {
        message(paste("Shapiro-Wilk检验（", group_label, "）失败：", e$message, sep = ""))
        return(NULL)
      }
    )
    if (!is.null(tmp)) {
      res <- tibble::add_row(res,
                             Group = group_label,
                             Test = "Shapiro-Wilk",
                             Statistic = tmp$statistic,
                             p.value = tmp$p.value)
    }
  } else {
    res <- tibble::add_row(res,
                           Group = group_label,
                           Test = "Shapiro-Wilk",
                           Statistic = NA_real_,
                           p.value = NA_real_)
    message(paste("Shapiro-Wilk检验（", group_label, "）跳过：样本量需3-5000（当前n=", n, "）", sep = ""))
  }
  
  # 2. Lilliefors-KS检验
  if (requireNamespace("nortest", quietly = TRUE)) {
    tmp <- tryCatch(
      expr = nortest::lillie.test(x),
      error = function(e) {
        message(paste("Lilliefors-KS检验（", group_label, "）失败：", e$message, sep = ""))
        return(NULL)
      }
    )
    if (!is.null(tmp)) {
      res <- tibble::add_row(res,
                             Group = group_label,
                             Test = "Lilliefors-KS",
                             Statistic = tmp$statistic,
                             p.value = tmp$p.value)
    } else {
      res <- tibble::add_row(res,
                             Group = group_label,
                             Test = "Lilliefors-KS",
                             Statistic = NA_real_,
                             p.value = NA_real_)
    }
  } else {
    res <- tibble::add_row(res,
                           Group = group_label,
                           Test = "Lilliefors-KS",
                           Statistic = NA_real_,
                           p.value = NA_real_)
    message("提示：需安装nortest包以运行KS/AD检验")
  }
  
  # 3. Anderson-Darling检验
  if (requireNamespace("nortest", quietly = TRUE)) {
    tmp <- tryCatch(
      expr = nortest::ad.test(x),
      error = function(e) {
        message(paste("Anderson-Darling检验（", group_label, "）失败：", e$message, sep = ""))
        return(NULL)
      }
    )
    if (!is.null(tmp)) {
      res <- tibble::add_row(res,
                             Group = group_label,
                             Test = "Anderson-Darling",
                             Statistic = tmp$statistic,
                             p.value = tmp$p.value)
    } else {
      res <- tibble::add_row(res,
                             Group = group_label,
                             Test = "Anderson-Darling",
                             Statistic = NA_real_,
                             p.value = NA_real_)
    }
  } else {
    res <- tibble::add_row(res,
                           Group = group_label,
                           Test = "Anderson-Darling",
                           Statistic = NA_real_,
                           p.value = NA_real_)
  }
  
  res
}

# ------------------------------
# Summary方法
# ------------------------------
#' @export
summary.mda <- function(object, dec = 3, ...) {
  # 1. 基础信息
  cat("Multigroup Difference Analysis (ANOVA/KW)\n")
  cat("Data         :", object$df_name, "\n")
  cat("Dependent var:", object$var, "(numeric)\n")
  cat("Group var    :", object$group, "(factor,", length(levels(object$plot_obj$homo$data[[object$group]])), "levels)\n")
  cat("Normality test:", object$normality_type, "\n")
  cat("Valid samples:", object$valid_n, "\n\n")
  
  # 2. 正态性检验结果
  cat("=== 1. Normality Test Results ===\n")
  if (nrow(object$norm_res) == 0) {
    cat("  No valid normality test results.\n\n")
  } else {
    norm_formatted <- object$norm_res %>%
      dplyr::mutate(
        Statistic = as.character(round(Statistic, dec)),  # 转为字符型，统一类型
        p.value = dplyr::case_when(
          is.na(p.value) ~ "",
          p.value < 0.001 ~ "<0.001",
          p.value < 0.01 ~ as.character(round(p.value, 3)),  # 数值转字符
          TRUE ~ as.character(round(p.value, 4))  # 数值转字符
        )
      ) %>%
      as.data.frame(stringsAsFactors = FALSE)
    
    print(norm_formatted, row.names = FALSE, right = FALSE)
    cat("\n")
  }
  
  # 3. 方差齐性检验结果
  cat("=== 2. Homogeneity of Variance Results ===\n")
  if (nrow(object$homo_res) == 0) {
    cat("  No valid homogeneity test results.\n\n")
  } else {
    homo_formatted <- object$homo_res %>%
      dplyr::mutate(
        Statistic = as.character(round(Statistic, dec)),  # 转为字符型，统一类型
        p.value = dplyr::case_when(
          is.na(p.value) ~ "",
          p.value < 0.001 ~ "<0.001",
          p.value < 0.01 ~ as.character(round(p.value, 3)),  # 数值转字符
          TRUE ~ as.character(round(p.value, 4))  # 数值转字符
        )
      ) %>%
      as.data.frame(stringsAsFactors = FALSE)
    
    print(homo_formatted, row.names = FALSE, right = FALSE)
    cat("\n")
  }
  
  # 4. 结论提示
  cat("=== 3. Interpretation Tips ===\n")
  cat("• 正态性：p ≥ 0.05 → 满足正态性假设\n")
  cat("• 方差齐性：p ≥ 0.05 → 满足方差齐性假设\n")
  cat("• 若同时满足这两个假设 → 使用方差分析（ANOVA）\n")
  cat("• 若任一假设不满足 → 使用Kruskal-Wallis检验\n")
  
  invisible(object)
}

# ------------------------------
# Plot方法
# ------------------------------
#' @export
plot.mda <- function(x,
                     plots = c("norm_qq", "norm_hist", "homo_box"),
                     shiny = FALSE, custom = FALSE, ...) {
  # 1. 基础校验
  if (length(plots) == 0) {
    return(ggplot2::ggplot() +
             ggplot2::annotate("text", x = 1, y = 1, label = i18n$t("No plots selected")) +
             ggplot2::theme_void())
  }
  
  plot_list <- list()
  var_name <- x$var
  group_name <- x$group
  
  # 2. 正态性检验图表
  # 2.1 Q-Q图
  if ("norm_qq" %in% plots) {
    if (x$normality_type == "overall") {
      p <- ggplot2::ggplot(data.frame(y = x$plot_obj$norm$data), ggplot2::aes(sample = y)) +
        ggplot2::stat_qq(color = "#2E86AB", size = 1) +
        ggplot2::stat_qq_line(color = "#A23B72", linetype = "dashed") +
        ggplot2::labs(x = "Theoretical Quantiles",
                      y = paste("Empirical Quantiles (", var_name, ")", sep = ""),
                      title = "Normality: Q-Q Plot (Overall)") +
        ggplot2::theme_minimal() +
        ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5, size = 12))
      plot_list[["norm_qq"]] <- p
    } else {
      # 按分组画QQ图
      group_data <- x$plot_obj$norm$group_data
      for (level in names(group_data)) {
        p <- ggplot2::ggplot(data.frame(y = group_data[[level]]), ggplot2::aes(sample = y)) +
          ggplot2::stat_qq(color = "#2E86AB", size = 1) +
          ggplot2::stat_qq_line(color = "#A23B72", linetype = "dashed") +
          ggplot2::labs(x = "Theoretical Quantiles",
                        y = paste("Empirical Quantiles (", var_name, ")", sep = ""),
                        title = paste("Normality: Q-Q Plot (", level, ")", sep = "")) +
          ggplot2::theme_minimal() +
          ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5, size = 12))
        plot_list[[paste("norm_qq_", level, sep = "")]] <- p
      }
    }
  }
  
  # 2.2 直方图
  if ("norm_hist" %in% plots) {
    if (x$normality_type == "overall") {
      p <- ggplot2::ggplot(data.frame(y = x$plot_obj$norm$data), ggplot2::aes(x = y)) +
        ggplot2::geom_histogram(fill = "#F18F01", alpha = 0.7, bins = 30) +
        ggplot2::labs(x = var_name, y = "Count",
                      title = "Normality: Histogram (Overall)") +
        ggplot2::theme_minimal() +
        ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5, size = 12))
      plot_list[["norm_hist"]] <- p
    } else {
      # 按分组画直方图
      group_data <- x$plot_obj$norm$group_data
      for (level in names(group_data)) {
        p <- ggplot2::ggplot(data.frame(y = group_data[[level]]), ggplot2::aes(x = y)) +
          ggplot2::geom_histogram(fill = "#F18F01", alpha = 0.7, bins = 30) +
          ggplot2::labs(x = var_name, y = "Count",
                        title = paste("Normality: Histogram (", level, ")", sep = "")) +
          ggplot2::theme_minimal() +
          ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5, size = 12))
        plot_list[[paste("norm_hist_", level, sep = "")]] <- p
      }
    }
  }
  
  # 3. 方差齐性检验图表
  if ("homo_box" %in% plots) {
    p <- ggplot2::ggplot(x$plot_obj$homo$data,
                         ggplot2::aes(x = .data[[group_name]],
                                      y = .data[[var_name]],
                                      fill = .data[[group_name]])) +
      ggplot2::geom_boxplot(alpha = 0.7, show.legend = FALSE) +
      ggplot2::scale_fill_brewer(palette = "Set2") +
      ggplot2::labs(x = group_name, y = var_name,
                    title = "Homogeneity: Boxplot by Group") +
      ggplot2::theme_minimal() +
      ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5, size = 12))
    plot_list[["homo_box"]] <- p
  }
  
  # 4. 组合图表
  combined_plot <- patchwork::wrap_plots(plot_list, ncol = 1, guides = "collect")
  
  # 5. 输出
  if (shiny) {
    print(combined_plot)
    return(invisible(combined_plot))  
  } else {
    return(combined_plot) 
  }
}