Skip to content
TheChatArchiveThe Chat Archive
Claudeclaude-sonnet-4-20250514-thinking-32k· AnthropicHugging Face LMArena Human Preference 140K

library(igraph) library(ggplot2) library(dplyr) library(purrr) library(visNetwork)…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Metallama-4-scout-17b-16e-instruct’s answer.
  1. User

    library(igraph) library(ggplot2) library(dplyr) library(purrr) library(visNetwork) library(ggraph) library(tidygraph) library(CCA) # do analizy kanonicznej

    ========================================

    🔹 Generowanie danych

    ========================================

    set.seed(123) n <- 100

    X1 <- matrix(rnorm(n*3), ncol=3) colnames(X1) <- c("X1_1", "X1_2", "X1_3")

    X2 <- matrix(rnorm(n*3), ncol=3) colnames(X2) <- c("X2_1", "X2_2", "X2_3")

    X3 <- matrix(rnorm(n*3), ncol=3) colnames(X3) <- c("X3_1", "X3_2", "X3_3")

    ========================================

    🔹 Obliczanie kanonicznych korelacji i ładunków

    ========================================

    Funkcja do obliczenia kanonicznej korelacji i ładunków między dwoma blokami

    cca_two_blocks <- function(Xa, Xb) { cca_res <- cc(Xa, Xb) # analiza kanoniczna list( cor = cca_rescor, # korelacje kanoniczne loadings_A = cca_resxcoef, # ładunki kanoniczne dla Xa loadings_B = cca_resycoef, # ładunki kanoniczne dla Xb cross_loadings_A = cor(Xa, cca_resscoresxscores), # ładunki krzyżowe dla Xa cross_loadings_B = cor(Xb, cca_resscores$yscores) # ładunki krzyżowe dla Xb ) }

    Liczymy kanoniczne korelacje i ładunki dla par bloków

    cca_12 <- cca_two_blocks(X1, X2) cca_23 <- cca_two_blocks(X2, X3) cca_13 <- cca_two_blocks(X1, X3)

    Korelacje między zbiorami - weźmy maksymalną korelację kanoniczną z pierwszego wymiaru

    cor_matrix <- matrix(c( 1, cca_12cor[1],cca13cor[1], cca_13cor[1], cca_12cor[1],1,cca23cor[1], 1, cca_23cor[1], cca_13cor[1],cca23cor[1], cca_23cor[1], 1 ), nrow=3, byrow=TRUE)

    Ładunki kanoniczne na pierwszy wymiar (uśrednione po zmiennych każdego zbioru, absolutne wartości)

    loadings <- list( list(as.vector(abs(cca_12loadingsA[,1])),as.vector(abs(cca12loadings_A[,1])), as.vector(abs(cca_12loadings_B[,1]))), # 1 wymiar: X1 i X2 list(as.vector(abs(cca_23loadingsA[,1])),as.vector(abs(cca23loadings_A[,1])), as.vector(abs(cca_23loadings_B[,1]))), # 1 wymiar: X2 i X3 list(as.vector(abs(cca_13loadingsA[,1])),as.vector(abs(cca13loadings_A[,1])), as.vector(abs(cca_13loadings_B[,1]))) # 1 wymiar: X1 i X3 )

    Ponieważ mamy 3 zbiory, bierzemy ładunki kanoniczne z każdego bloku

    Aby nadać je zmiennym w zbiorach, ułożymy je tak:

    loadings_list <- list( setNames(abs(cca_12loadingsA[,1]),colnames(X1)),setNames(abs(cca12loadings_A[,1]), colnames(X1)), setNames(abs(cca_12loadings_B[,1]), colnames(X2)), setNames(abs(cca_23$loadings_B[,1]), colnames(X3)) )

    ========================================

    🔹 Tworzenie węzłów i krawędzi (bez zmian)

    ========================================

    nodes <- data.frame( name = c(paste("Zbiór", 1:3), colnames(X1), colnames(X2), colnames(X3)), type = c(rep("set", 3), rep("var", length(c(colnames(X1), colnames(X2), colnames(X3))))) )

    Krawędzie między zbiorami

    edges_sets <- data.frame( from = c("Zbiór 1", "Zbiór 2", "Zbiór 1"), to = c("Zbiór 2", "Zbiór 3", "Zbiór 3"), weight = c(cor_matrix[1, 2], cor_matrix[2, 3], cor_matrix[1,3]), Wymiar = 1 )

    Krawędzie zbiór → cecha z odpowiednimi ładunkami

    edges_vars <- data.frame( from = c(rep("Zbiór 1", length(X1[1,])), rep("Zbiór 2", length(X2[1,])), rep("Zbiór 3", length(X3[1,]))), to = c(colnames(X1), colnames(X2), colnames(X3)), weight = c(loadings_list[[1]], loadings_list[[2]], loadings_list[[3]]), Wymiar = 1 )

    edges <- rbind(edges_sets, edges_vars) g <- graph_from_data_frame(edges, vertices = nodes, directed = FALSE)

    ========================================

    🔹 Etykiety i style (tak jak w Twoim kodzie)

    ========================================

    V(g)label<−sapply(V(g)label <- sapply(V(g)name, function(n) { if (n %in% paste("Zbiór", 1:3)) n else paste0(n, "\n(", round(unlist(loadings_list[which(sapply(loadings_list, function(x) n %in% names(x)))][[1]][n]), 2), ")") })

    V(g)size<−ifelse(V(g)size <- ifelse(V(g)type == "set", 50, 35) V(g)label.cex<−ifelse(V(g)label.cex <- ifelse(V(g)type == "set", 1.1, 0.85) V(g)$frame.color <- "black"

    E(g)color<−ifelse(edgescolor <- ifelse( edgesfrom %in% paste("Zbiór", 1:3) & edgesto"red","gray")E(g)to %in% paste("Zbiór", 1:3), "red", "gray" ) E(g)width <- ifelse(E(g)color=="red",3,1.5)E(g)color == "red", 3, 1.5) E(g)curved <- ifelse(E(g)color=="red",0.2,0)edgelabels<−ifelse(E(g)color == "red", 0.2, 0) edge_labels <- ifelse(E(g)color == "red", round(E(g)$weight, 2), "")

    ========================================

    🎨 Kolorowanie zmiennych wg progu

    ========================================

    threshold <- 0.7

    node_colors <- rep("lightgreen", length(V(g))) for (i in seq_along(V(g))) { if (V(g)type[i] == "var") { varname <- V(g)name[i] if (abs(unlist(loadings_list[which(sapply(loadings_list, function(x) varname %in% names(x)))][[1]][varname])) >= threshold) { node_colors[i] <- "forestgreen" } } } node_colors[V(g)type=="set"]<−"lightblue"V(g)type == "set"] <- "lightblue" V(g)color <- node_colors

    ========================================

    📐 Układ - 3 sektory ("koło")

    ========================================

    layout_mat <- matrix(NA, nrow = length(V(g)), ncol = 2) rownames(layout_mat) <- V(g)$name

    Pozycje zbiorów

    radius <- 3 angles_deg <- c(210, 90, 330) angles_rad <- angles_deg * pi/180 set_names <- paste("Zbiór", 1:3)

    for (i in 1:3) { layout_mat[set_names[i], ] <- c( radius * cos(angles_rad[i]), radius * sin(angles_rad[i]) ) }

    Funkcja rozmieszczania cech względem zbioru

    place_vars_sector <- function(set_name, var_names, center_angle_rad, spread = pi/4, min_dist = 2, max_extra = 2) { set_pos <- layout_mat[set_name, ] n <- length(var_names) angles <- seq(center_angle_rad - spread, center_angle_rad + spread, length.out = n)

    for (i in seq_along(var_names)) { var <- var_names[i] loading <- abs(unlist(loadings_list[which(sapply(loadings_list, function(x) var %in% names(x)))][[1]][var])) dist <- min_dist + (1 - loading) * max_extra layout_mat[var, ] <<- set_pos + dist * c(cos(angles[i]), sin(angles[i])) } }

    Rozkład cech względem zbiorów

    place_vars_sector("Zbiór 1", colnames(X1), angles_rad[1]) place_vars_sector("Zbiór 2", colnames(X2), angles_rad[2]) place_vars_sector("Zbiór 3", colnames(X3), angles_rad[3])

    ========================================

    🖼️ Rysowanie wykresu

    ========================================

    par(mar = c(1, 1, 4, 1)) plot( g, layout = layout_mat, edge.label = edge_labels, edge.label.family = "sans", edge.label.color = "blue", edge.label.cex = 1.3, main = "Struktura powiązań między zbiorami i ich zmiennymi (Wymiar 1)" )

    Legenda

    legend("bottomright", legend = c("Zbiory", "Zmienne (ładunek < 0.7)", "Zmienne (ładunek ≥ 0.7)", "Korelacje między zbiorami", "Ładunki kanoniczne"), pch = c(19, 19, 19, NA, NA), lty = c(NA, NA, NA, 1, 1), col = c("lightblue", "lightgreen", "forestgreen", "red", "gray"), pt.cex = 2, bty = "n", cex = 0.85)

    Podpis informacyjny

    mtext(paste0( "Średnia korelacja między zbiorami: ", round(mean(cor_matrix[upper.tri(cor_matrix)]), 2), "\nŚrednia siła ładunków kanonicznych: ", round(mean(unlist(loadings_list)), 2) ), side = 1, line = -2, cex = 0.8)

    ################################################################################ ################################################################################

    library(igraph) library(igraph) library(ggplot2) library(dplyr) library(purrr) library(tidyr) library(ggraph) library(tidygraph) library(visNetwork)

    set.seed(123)

    === Generowanie 3 zbiorów po 3 zmienne ===

    n <- 50 X1 <- data.frame(X1_1 = rnorm(n), X1_2 = rnorm(n), X1_3 = rnorm(n)) X2 <- data.frame(X2_1 = rnorm(n), X2_2 = rnorm(n), X2_3 = rnorm(n)) X3 <- data.frame(X3_1 = rnorm(n), X3_2 = rnorm(n), X3_3 = rnorm(n))

    === Ulepszona funkcja do analizy kanonicznej ===

    compute_extended_cancor <- function(A, B) {

    Standaryzacja danych

    A_std <- scale(A) B_std <- scale(B)

    Analiza kanoniczna

    cc <- cancor(A_std, B_std)

    Zmienne kanoniczne

    U <- as.matrix(A_std) %*% ccxcoefV<−as.matrix(Bstd)xcoef V <- as.matrix(B_std) %*% ccycoef

    Ładunki strukturalne

    struct_A <- cor(A_std, U) struct_B <- cor(B_std, V)

    Ładunki krzyżowe

    cross_A <- cor(A_std, V) cross_B <- cor(B_std, U)

    Test Wilksa dla istotności korelacji kanonicznych

    n <- nrow(A) p <- ncol(A) q <- ncol(B)

    wilks_lambda <- prod(1 - cc$cor^2) chi_sq <- -(n - 1 - (p + q + 1)/2) * log(wilks_lambda) df <- p * q p_value <- 1 - pchisq(chi_sq, df)

    list( correlations = cccor,loadingsA=cccor, loadings_A = ccxcoef, loadings_B = ccycoef,structuralA=structA,structuralB=structB,crossA=crossA,crossB=crossB,canonicalR2=ccycoef, structural_A = struct_A, structural_B = struct_B, cross_A = cross_A, cross_B = cross_B, canonical_R2 = cccor^2, wilks_test = list( lambda = wilks_lambda, chi_sq = chi_sq, df = df, p_value = p_value ) ) }

    norm_loadings <- function(loadings) { loadings_matrix <- as.matrix(loadings) result <- apply(loadings_matrix, 2, function(col) { max_abs <- max(abs(col)) ifelse(max_abs > 0, col / max_abs, col) })

    Upewniamy się, że wynik jest macierzą

    if (is.vector(result)) { result <- matrix(result, ncol = 1) rownames(result) <- rownames(loadings_matrix) colnames(result) <- colnames(loadings_matrix)[1] } return(result) }

    === Obliczenia dla wszystkich par ===

    res_12 <- compute_extended_cancor(X1, X2) res_23 <- compute_extended_cancor(X2, X3) res_13 <- compute_extended_cancor(X1, X3)

    === Przygotowanie znormalizowanych ładunków strukturalnych ===

    load_12_A <- norm_loadings(as.matrix(res_12structuralA))load12B<−normloadings(as.matrix(res12structural_A)) load_12_B <- norm_loadings(as.matrix(res_12structural_B)) load_23_A <- norm_loadings(as.matrix(res_23structuralA))load23B<−normloadings(as.matrix(res23structural_A)) load_23_B <- norm_loadings(as.matrix(res_23structural_B)) load_13_A <- norm_loadings(as.matrix(res_13structuralA))load13B<−normloadings(as.matrix(res13structural_A)) load_13_B <- norm_loadings(as.matrix(res_13structural_B))

    Ładunki w liście po 3 zbiorach i 2 wymiarach

    loadings <- list( list( load_12_A[,1], load_12_A[,2] ), list( load_12_B[,1], load_12_B[,2] ), list( load_23_B[,1], load_23_B[,2] ) )

    names(loadings[[1]][[1]]) <- colnames(X1) names(loadings[[1]][[2]]) <- colnames(X1) names(loadings[[2]][[1]]) <- colnames(X2) names(loadings[[2]][[2]]) <- colnames(X2) names(loadings[[3]][[1]]) <- colnames(X3) names(loadings[[3]][[2]]) <- colnames(X3)

    === Funkcja do wyświetlania podsumowania ===

    print_summary <- function(res_12, res_23, res_13) { cat("\n=== Podsumowanie analizy kanonicznej ===\n")

    cat("\nKorelacje kanoniczne:\n") cat("X1-X2:", round(res_12correlations,3),"\nIstotnosˊcˊ(p−value):",round(res12correlations, 3), "\nIstotność (p-value):", round(res_12wilks_testpvalue,3),"\n")cat("X2−X3:",round(res23p_value, 3), "\n") cat("X2-X3:", round(res_23correlations, 3), "\nIstotność (p-value):", round(res_23wilkstestwilks_testp_value, 3), "\n") cat("X1-X3:", round(res_13correlations,3),"\nIstotnosˊcˊ(p−value):",round(res13correlations, 3), "\nIstotność (p-value):", round(res_13wilks_test$p_value, 3), "\n")

    cat("\nWyjaśniona wariancja (R²):\n") cat("X1-X2:", round(res_12canonicalR2,3),"\n")cat("X2−X3:",round(res23canonical_R2, 3), "\n") cat("X2-X3:", round(res_23canonical_R2, 3), "\n") cat("X1-X3:", round(res_13$canonical_R2, 3), "\n") }

    === Funkcja do rysowania grafu ===

    draw_improved_graph <- function(dim_num, res_12, res_23, res_13) { load_dim <- map(loadings, ~ .x[[dim_num]])

    cor_matrix_dim1 <- matrix(c( 1, mean(c(res_12correlations[1],res13correlations[1], res_13correlations[1])), mean(c(res_12correlations[1],res23correlations[1], res_23correlations[1])), mean(c(res_12correlations[1],res13correlations[1], res_13correlations[1])), 1, mean(c(res_12correlations[1],res23correlations[1], res_23correlations[1])), mean(c(res_12correlations[1],res23correlations[1], res_23correlations[1])), mean(c(res_12correlations[1],res23correlations[1], res_23correlations[1])), 1 ), nrow = 3, byrow = TRUE)

    cor_matrix_dim2 <- matrix(c( 1, mean(c(res_12correlations[2],res13correlations[2], res_13correlations[2])), mean(c(res_12correlations[2],res23correlations[2], res_23correlations[2])), mean(c(res_12correlations[2],res13correlations[2], res_13correlations[2])), 1, mean(c(res_12correlations[2],res23correlations[2], res_23correlations[2])), mean(c(res_12correlations[2],res23correlations[2], res_23correlations[2])), mean(c(res_12correlations[2],res23correlations[2], res_23correlations[2])), 1 ), nrow = 3, byrow = TRUE)

    cor_mat <- if(dim_num == 1) cor_matrix_dim1 else cor_matrix_dim2

    nodes <- data.frame( name = c(paste("Zbiór", 1:3), colnames(X1), colnames(X2), colnames(X3)), type = c(rep("set", 3), rep("var", 9)) )

    loading_values <- list( setNames(load_dim[[1]], colnames(X1)), setNames(load_dim[[2]], colnames(X2)), setNames(load_dim[[3]], colnames(X3)) )

    edges_sets <- data.frame( from = c("Zbiór 1", "Zbiór 2", "Zbiór 1"), to = c("Zbiór 2", "Zbiór 3", "Zbiór 3"), weight = c(cor_mat[1, 2], cor_mat[2, 3], cor_mat[1,3]), Wymiar = dim_num )

    edges_vars <- data.frame( from = c(rep("Zbiór 1", 3), rep("Zbiór 2", 3), rep("Zbiór 3", 3)), to = c(colnames(X1), colnames(X2), colnames(X3)), weight = c(abs(load_dim[[1]]), abs(load_dim[[2]]), abs(load_dim[[3]])), Wymiar = dim_num )

    edges <- rbind(edges_sets, edges_vars)

    g <- graph_from_data_frame(edges, vertices = nodes, directed = FALSE)

    V(g)label<−sapply(V(g)label <- sapply(V(g)name, function(n) { if (n %in% paste("Zbiór", 1:3)) n else { for (lv in loading_values) { if (n %in% names(lv)) return(paste0(n, "\n(", round(lv[n], 2), ")")) } n } })

    V(g)size<−ifelse(V(g)size <- ifelse(V(g)type == "set", 50, 35) V(g)label.cex<−ifelse(V(g)label.cex <- ifelse(V(g)type == "set", 1.1, 0.85) V(g)$frame.color <- "black"

    E(g)color<−ifelse(edgescolor <- ifelse( edgesfrom %in% paste("Zbiór", 1:3) & edgesto"red","gray")E(g)to %in% paste("Zbiór", 1:3), "red", "gray" ) E(g)width <- ifelse(E(g)color=="red",3,1.5)E(g)color == "red", 3, 1.5) E(g)curved <- ifelse(E(g)color=="red",0.2,0)edgelabels<−ifelse(E(g)color == "red", 0.2, 0) edge_labels <- ifelse(E(g)color == "red", round(E(g)$weight, 2), "")

    threshold <- 0.7 node_colors <- rep("lightgreen", length(V(g))) for (i in seq_along(V(g))) { if (V(g)type[i] == "var") { varname <- V(g)name[i] if (abs(unlist(loading_values[which(sapply(loading_values, function(x) varname %in% names(x)))][[1]][varname])) >= threshold) { node_colors[i] <- "forestgreen" } } } node_colors[V(g)type=="set"]<−"lightblue"V(g)type == "set"] <- "lightblue" V(g)color <- node_colors

    layout_mat <- matrix(NA, nrow = length(V(g)), ncol = 2) rownames(layout_mat) <- V(g)$name radius <- 3 angles_deg <- c(210, 90, 330) angles_rad <- angles_deg * pi/180 set_names <- paste("Zbiór", 1:3)

    for (i in 1:3) { layout_mat[set_names[i], ] <- c(radius * cos(angles_rad[i]), radius * sin(angles_rad[i])) }

    place_vars_sector <- function(set_name, var_names, center_angle_rad, spread = pi/4, min_dist = 2, max_extra = 2) { set_pos <- layout_mat[set_name, ] n <- length(var_names) angles <- seq(center_angle_rad - spread, center_angle_rad + spread, length.out = n)

    for (i in seq_along(var_names)) {
      var <- var_names[i]
      loading <- abs(unlist(loading_values[which(sapply(loading_values, function(x) var %in% names(x)))][[1]][var]))
      
         popraw ten kod, ma mieć wszystko
    
  2. Claude Assistant

    Oto poprawiony i kompletny kod:

    library(igraph)
    library(ggplot2)
    library(dplyr)
    library(purrr)
    library(tidyr)
    library(ggraph)
    library(tidygraph)
    library(visNetwork)
    
    set.seed(123)
    
    # === Generowanie 3 zbiorów po 3 zmienne ===
    n <- 50
    X1 <- data.frame(X1_1 = rnorm(n), X1_2 = rnorm(n), X1_3 = rnorm(n))
    X2 <- data.frame(X2_1 = rnorm(n), X2_2 = rnorm(n), X2_3 = rnorm(n))
    X3 <- data.frame(X3_1 = rnorm(n), X3_2 = rnorm(n), X3_3 = rnorm(n))
    
    # === Ulepszona funkcja do analizy kanonicznej ===
    compute_extended_cancor <- function(A, B) {
      # Standaryzacja danych
      A_std <- scale(A)
      B_std <- scale(B)
      
      # Analiza kanoniczna
      cc <- cancor(A_std, B_std)
      
      # Zmienne kanoniczne
      U <- as.matrix(A_std) %*% cc$xcoef
      V <- as.matrix(B_std) %*% cc$ycoef
      
      # Ładunki strukturalne
      struct_A <- cor(A_std, U)
      struct_B <- cor(B_std, V)
      
      # Ładunki krzyżowe
      cross_A <- cor(A_std, V)
      cross_B <- cor(B_std, U)
      
      # Test Wilksa dla istotności korelacji kanonicznych
      n <- nrow(A)
      p <- ncol(A)
      q <- ncol(B)
      
      wilks_lambda <- prod(1 - cc$cor^2)
      chi_sq <- -(n - 1 - (p + q + 1)/2) * log(wilks_lambda)
      df <- p * q
      p_value <- 1 - pchisq(chi_sq, df)
      
      list(
        correlations = cc$cor,
        loadings_A = cc$xcoef,
        loadings_B = cc$ycoef,
        structural_A = struct_A,
        structural_B = struct_B,
        cross_A = cross_A,
        cross_B = cross_B,
        canonical_R2 = cc$cor^2,
        wilks_test = list(
          lambda = wilks_lambda,
          chi_sq = chi_sq,
          df = df,
          p_value = p_value
        )
      )
    }
    
    norm_loadings <- function(loadings) {
      loadings_matrix <- as.matrix(loadings)
      result <- apply(loadings_matrix, 2, function(col) {
        max_abs <- max(abs(col))
        ifelse(max_abs > 0, col / max_abs, col)
      })
      # Upewniamy się, że wynik jest macierzą
      if (is.vector(result)) {
        result <- matrix(result, ncol = 1)
        rownames(result) <- rownames(loadings_matrix)
        colnames(result) <- colnames(loadings_matrix)[1]
      }
      return(result)
    }
    
    # === Obliczenia dla wszystkich par ===
    res_12 <- compute_extended_cancor(X1, X2)
    res_23 <- compute_extended_cancor(X2, X3)
    res_13 <- compute_extended_cancor(X1, X3)
    
    # === Przygotowanie znormalizowanych ładunków strukturalnych ===
    load_12_A <- norm_loadings(as.matrix(res_12$structural_A))
    load_12_B <- norm_loadings(as.matrix(res_12$structural_B))
    load_23_A <- norm_loadings(as.matrix(res_23$structural_A))
    load_23_B <- norm_loadings(as.matrix(res_23$structural_B))
    load_13_A <- norm_loadings(as.matrix(res_13$structural_A))
    load_13_B <- norm_loadings(as.matrix(res_13$structural_B))
    
    # Ładunki w liście po 3 zbiorach i 2 wymiarach
    loadings <- list(
      list(
        load_12_A[,1], load_12_A[,2]
      ),
      list(
        load_12_B[,1], load_12_B[,2]
      ),
      list(
        load_23_B[,1], load_23_B[,2]
      )
    )
    
    names(loadings[[1]][[1]]) <- colnames(X1)
    names(loadings[[1]][[2]]) <- colnames(X1)
    names(loadings[[2]][[1]]) <- colnames(X2)
    names(loadings[[2]][[2]]) <- colnames(X2)
    names(loadings[[3]][[1]]) <- colnames(X3)
    names(loadings[[3]][[2]]) <- colnames(X3)
    
    # === Funkcja do wyświetlania podsumowania ===
    print_summary <- function(res_12, res_23, res_13) {
      cat("\n=== Podsumowanie analizy kanonicznej ===\n")
      
      cat("\nKorelacje kanoniczne:\n")
      cat("X1-X2:", round(res_12$correlations, 3), 
          "\nIstotność (p-value):", round(res_12$wilks_test$p_value, 3), "\n")
      cat("X2-X3:", round(res_23$correlations, 3),
          "\nIstotność (p-value):", round(res_23$wilks_test$p_value, 3), "\n")
      cat("X1-X3:", round(res_13$correlations, 3),
          "\nIstotność (p-value):", round(res_13$wilks_test$p_value, 3), "\n")
      
      cat("\nWyjaśniona wariancja (R²):\n")
      cat("X1-X2:", round(res_12$canonical_R2, 3), "\n")
      cat("X2-X3:", round(res_23$canonical_R2, 3), "\n")
      cat("X1-X3:", round(res_13$canonical_R2, 3), "\n")
    }
    
    # === Funkcja do rysowania grafu ===
    draw_improved_graph <- function(dim_num, res_12, res_23, res_13) {
      load_dim <- map(loadings, ~ .x[[dim_num]])
      
      cor_matrix_dim1 <- matrix(c(
        1,
        res_12$correlations[1],
        res_13$correlations[1],
        res_12$correlations[1],
        1,
        res_23$correlations[1],
        res_13$correlations[1],
        res_23$correlations[1],
        1
      ), nrow = 3, byrow = TRUE)
      
      cor_matrix_dim2 <- matrix(c(
        1,
        res_12$correlations[2],
        res_13$correlations[2],
        res_12$correlations[2],
        1,
        res_23$correlations[2],
        res_13$correlations[2],
        res_23$correlations[2],
        1
      ), nrow = 3, byrow = TRUE)
      
      cor_mat <- if(dim_num == 1) cor_matrix_dim1 else cor_matrix_dim2
      
      nodes <- data.frame(
        name = c(paste("Zbiór", 1:3), colnames(X1), colnames(X2), colnames(X3)),
        type = c(rep("set", 3), rep("var", 9))
      )
      
      loading_values <- list(
        setNames(load_dim[[1]], colnames(X1)),
        setNames(load_dim[[2]], colnames(X2)),
        setNames(load_dim[[3]], colnames(X3))
      )
      
      edges_sets <- data.frame(
        from = c("Zbiór 1", "Zbiór 2", "Zbiór 1"),
        to = c("Zbiór 2", "Zbiór 3", "Zbiór 3"),
        weight = c(cor_mat[1, 2], cor_mat[2, 3], cor_mat[1,3]),
        Wymiar = dim_num
      )
      
      edges_vars <- data.frame(
        from = c(rep("Zbiór 1", 3), rep("Zbiór 2", 3), rep("Zbiór 3", 3)),
        to = c(colnames(X1), colnames(X2), colnames(X3)),
        weight = c(abs(load_dim[[1]]), abs(load_dim[[2]]), abs(load_dim[[3]])),
        Wymiar = dim_num
      )
      
      edges <- rbind(edges_sets, edges_vars)
      
      g <- graph_from_data_frame(edges, vertices = nodes, directed = FALSE)
      
      V(g)$label <- sapply(V(g)$name, function(n) {
        if (n %in% paste("Zbiór", 1:3)) n else {
          for (lv in loading_values) {
            if (n %in% names(lv)) return(paste0(n, "\n(", round(lv[n], 2), ")"))
          }
          n
        }
      })
      
      V(g)$size <- ifelse(V(g)$type == "set", 50, 35)
      V(g)$label.cex <- ifelse(V(g)$type == "set", 1.1, 0.85)
      V(g)$frame.color <- "black"
      
      E(g)$color <- ifelse(
        edges$from %in% paste("Zbiór", 1:3) & edges$to %in% paste("Zbiór", 1:3),
        "red", "gray"
      )
      E(g)$width <- ifelse(E(g)$color == "red", 3, 1.5)
      E(g)$curved <- ifelse(E(g)$color == "red", 0.2, 0)
      edge_labels <- ifelse(E(g)$color == "red", round(E(g)$weight, 2), "")
      
      threshold <- 0.7
      node_colors <- rep("lightgreen", length(V(g)))
      for (i in seq_along(V(g))) {
        if (V(g)$type[i] == "var") {
          varname <- V(g)$name[i]
          for (lv in loading_values) {
            if (varname %in% names(lv)) {
              if (abs(lv[varname]) >= threshold) {
                node_colors[i] <- "forestgreen"
              }
              break
            }
          }
        }
      }
      node_colors[V(g)$type == "set"] <- "lightblue"
      V(g)$color <- node_colors
      
      # === Układanie węzłów ===
      layout_mat <- matrix(NA, nrow = length(V(g)), ncol = 2)
      rownames(layout_mat) <- V(g)$name
      radius <- 3
      angles_deg <- c(210, 90, 330)
      angles_rad <- angles_deg * pi/180
      set_names <- paste("Zbiór", 1:3)
      
      for (i in 1:3) {
        layout_mat[set_names[i], ] <- c(radius * cos(angles_rad[i]), radius * sin(angles_rad[i]))
      }
      
      place_vars_sector <- function(set_name, var_names, center_angle_rad, spread = pi/4, min_dist = 2, max_extra = 2) {
        set_pos <- layout_mat[set_name, ]
        n <- length(var_names)
        angles <- seq(center_angle_rad - spread, center_angle_rad + spread, length.out = n)
        
        for (i in seq_along(var_names)) {
          var <- var_names[i]
          loading <- 0
          for (lv in loading_values) {
            if (var %in% names(lv)) {
              loading <- abs(lv[var])
              break
            }
          }
          dist <- min_dist + (1 - loading) * max_extra
          layout_mat[var, ] <<- set_pos + dist * c(cos(angles[i]), sin(angles[i]))
        }
      }
      
      # Rozmieszczenie zmiennych względem zbiorów
      place_vars_sector("Zbiór 1", colnames(X1), angles_rad[1])
      place_vars_sector("Zbiór 2", colnames(X2), angles_rad[2])
      place_vars_sector("Zbiór 3", colnames(X3), angles_rad[3])
      
      # === Rysowanie wykresu ===
      par(mar = c(3, 1, 4, 1))
      plot(
        g,
        layout = layout_mat,
        edge.label = edge_labels,
        edge.label.family = "sans",
        edge.label.color = "blue",
        edge.label.cex = 1.3,
        main = paste("Struktura powiązań między zbiorami i ich zmiennymi (Wymiar", dim_num, ")")
      )
      
      # Legenda
      legend("bottomright",
             legend = c("Zbiory", "Zmienne (ładunek < 0.7)", "Zmienne (ładunek ≥ 0.7)",
                        "Korelacje między zbiorami", "Ładunki kanoniczne"),
             pch = c(19, 19, 19, NA, NA),
             lty = c(NA, NA, NA, 1, 1),
             col = c("lightblue", "lightgreen", "forestgreen", "red", "gray"),
             pt.cex = 2,
             bty = "n",
             cex = 0.85)
      
      # Podpis informacyjny
      avg_cor <- if(dim_num == 1) {
        mean(c(res_12$correlations[1], res_23$correlations[1], res_13$correlations[1]))
      } else {
        mean(c(res_12$correlations[2], res_23$correlations[2], res_13$correlations[2]))
      }
      
      avg_loading <- mean(abs(unlist(load_dim)))
      
      mtext(paste0(
        "Średnia korelacja między zbiorami: ", round(avg_cor, 2),
        "\nŚrednia siła ładunków kanonicznych: ", round(avg_loading, 2)
      ), side = 1, line = -1, cex = 0.8)
    }
    
    # === Wykonanie analizy i wizualizacji ===
    print_summary(res_12, res_23, res_13)
    
    # Rysowanie wykresów dla obu wymiarów
    par(mfrow = c(1, 2))
    draw_improved_graph(1, res_12, res_23, res_13)
    draw_improved_graph(2, res_12, res_23, res_13)
    

    Główne poprawki w kodzie:

    1. Dokończenie funkcji place_vars_sector - dodałem właściwe wyszukiwanie ładunków dla każdej zmiennej i obliczanie dystansu
    2. Poprawienie logiki kolorowania węzłów - naprawiłem pętlę sprawdzającą wartości ładunków
    3. Dodanie wywołań funkcji - dodałem print_summary() i wywołania draw_improved_graph() dla obu wymiarów
    4. Poprawienie obliczeń średnich - naprawiłem obliczanie średnich korelacji i ładunków dla podpisu
    5. Ustawienie układu wykresów - dodałem par(mfrow = c(1, 2)) do wyświetlania obu wymiarów obok siebie

    Kod teraz jest kompletny i powinien działać bez błędów, tworząc wizualizacje sieci dla obu wymiarów kanonicznych wraz z podsumowaniem statystycznym.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.