[HEAD]: initial commit

2026-08-04 19:19:24 +01:00
commit 2d9bfafbe6
39 changed files with 8905 additions and 0 deletions
@@ -0,0 +1,570 @@
+##panel regression----
+##all data with multicollinearity accounted for
+# panel_all <- plm(
+#   European_performance ~
+#     First_and_second +
+#     First_and_CL +
+#     CL_and_relegated +
+#     Top_4_total_GD +
+#     Average_number_of_players +
+#     Average_goals_by_top_3_players +
+#     Gini_coefficient,
+#   data = scaled_full_data,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_all)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_all, vcov = vcovHC, type = "HC1")
+
+# #rerun with only significant vars
+# panel_all_robust <- plm(
+#   European_performance ~
+#     Top_4_total_GD +
+#     Average_goals_by_top_3_players +
+#     Gini_coefficient,
+#   data = scaled_full_data,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_all_robust)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_all_robust, vcov = vcovHC, type = "HC1")
+
+#rerun with only significant vars & theory included
+panel_all_robust_theory <- plm(
+  European_performance ~
+    First_and_second +
+    CL_and_relegated +
+    Top_4_total_GD +
+    Average_goals_by_top_3_players +
+    Gini_coefficient,
+  data = scaled_full_data,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_all_robust_theory)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_all_robust_theory, vcov = vcovHC, type = "HC1")
+
+
+##using principal components
+panel_pc <- plm(
+  European_performance ~
+    PC1 +
+    PC2 +
+    PC3 +
+    PC4 +
+    PC5,
+  data = pca_results,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_pc)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_pc, vcov = vcovHC, type = "HC1")
+
+# #rerun with only significant pcs
+# panel_pc_robust <- plm(
+#   European_performance ~
+#     PC1 +
+#     PC4,
+#   data = pca_results,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_pc_robust)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_pc_robust, vcov = vcovHC, type = "HC1")
+
+
+##using competitive indices
+panel_indices_multi <- plm(
+  European_performance ~
+    competitive_title_index +
+    elite_dominance_index +
+    player_rotation_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_multi)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_multi, vcov = vcovHC, type = "HC1")
+
+panel_indices_title <- plm(
+  European_performance ~ competitive_title_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_title)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_title, vcov = vcovHC, type = "HC1")
+
+panel_indices_dominance <- plm(
+  European_performance ~ elite_dominance_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_dominance)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_dominance, vcov = vcovHC, type = "HC1")
+
+panel_indices_competitiveness <- plm(
+  European_performance ~ competitive_balance_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_competitiveness)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_competitiveness, vcov = vcovHC, type = "HC1")
+
+panel_indices_points <- plm(
+  European_performance ~ multilevel_competitiveness_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_points)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_points, vcov = vcovHC, type = "HC1")
+
+# panel_indices_players <- plm(
+#   European_performance ~ player_rotation_index,
+#   data = scaled_full_data_competitiveness_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_players)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_players, vcov = vcovHC, type = "HC1")
+
+indices_all_plotting_data <- regression_coefficient_data(panel_indices_multi) %>%
+  mutate(term = paste0(term, " (multi)")) %>%
+  rbind(regression_coefficient_data(panel_indices_title)) %>%
+  rbind(regression_coefficient_data(panel_indices_dominance)) %>%
+  rbind(regression_coefficient_data(panel_indices_competitiveness)) %>%
+  rbind(regression_coefficient_data(panel_indices_points))
+
+
+##panel regression using next years European performance value----
+##all data with multicollinearity accounted for
+panel_all_lead <- plm(
+  lead_european_performance ~
+    First_and_second +
+    First_and_CL +
+    CL_and_relegated +
+    Top_4_total_GD +
+    Average_number_of_players +
+    Average_goals_by_top_3_players +
+    Gini_coefficient,
+  data = scaled_full_data,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_all_lead)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_all_lead, vcov = vcovHC, type = "HC1")
+
+# #rerun with only significant vars & theory included
+# panel_all_lead_robust_theory <- plm(
+#   lead_european_performance ~
+#     First_and_CL +
+#     CL_and_relegated +
+#     Top_4_total_GD +
+#     Average_goals_by_top_3_players +
+#     Gini_coefficient,
+#   data = scaled_full_data,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_all_lead_robust_theory)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_all_lead_robust_theory, vcov = vcovHC, type = "HC1")
+
+
+##using principal components
+panel_lead_pc <- plm(
+  lead_european_performance ~
+    PC1 +
+    PC2 +
+    PC3 +
+    PC4 +
+    PC5,
+  data = pca_results,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_lead_pc)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_lead_pc, vcov = vcovHC, type = "HC1")
+
+#rerun with only significant pcs
+# panel_lead_pc_robust <- plm(
+#   lead_european_performance ~
+#     PC1 +
+#     PC3 +
+#     PC4,
+#   data = pca_results,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_lead_pc_robust)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_lead_pc_robust, vcov = vcovHC, type = "HC1")
+
+
+##using competitive indices
+panel_indices_multi_lead <- plm(
+  lead_european_performance ~
+    competitive_title_index +
+    elite_dominance_index +
+    player_rotation_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_multi_lead)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_multi_lead, vcov = vcovHC, type = "HC1")
+
+# panel_indices_title_lead <- plm(
+#   lead_european_performance ~ competitive_title_index,
+#   data = scaled_full_data_competitiveness_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_title_lead)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_title_lead, vcov = vcovHC, type = "HC1")
+
+panel_indices_dominance_lead <- plm(
+  lead_european_performance ~ elite_dominance_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_dominance_lead)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_dominance_lead, vcov = vcovHC, type = "HC1")
+
+panel_indices_competitiveness_lead <- plm(
+  lead_european_performance ~ competitive_balance_index,
+  data = scaled_full_data_competitiveness_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_competitiveness_lead)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_competitiveness_lead, vcov = vcovHC, type = "HC1")
+
+# panel_indices_points_lead <- plm(
+#   lead_european_performance ~ multilevel_competitiveness_index,
+#   data = scaled_full_data_competitiveness_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_points_lead)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_points_lead, vcov = vcovHC, type = "HC1")
+
+# panel_indices_players_lead <- plm(
+#   lead_european_performance ~ player_rotation_index,
+#   data = scaled_full_data_competitiveness_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_players_lead)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_players_lead, vcov = vcovHC, type = "HC1")
+
+indices_lead_plotting_data <- regression_coefficient_data(panel_indices_multi_lead) %>%
+  mutate(term = paste0(term, " (multi)")) %>%
+  rbind(regression_coefficient_data(panel_indices_dominance_lead)) %>%
+  rbind(regression_coefficient_data(panel_indices_competitiveness_lead))
+
+
+##panel regression for CL----
+##all data with multicollinearity accounted for
+CL_scaled_data <- scaled_full_data_comp_breakdown %>% filter(competition == "CL")
+panel_all_CL <- plm(
+  European_performance ~
+    First_and_second +
+    First_and_CL +
+    CL_and_relegated +
+    Top_4_total_GD +
+    Average_number_of_players +
+    Average_goals_by_top_3_players +
+    Gini_coefficient,
+  data = CL_scaled_data,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_all_CL)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_all_CL, vcov = vcovHC, type = "HC1")
+
+#rerun with only significant vars
+# panel_all_CL_robust <- plm(
+#   European_performance ~
+#     CL_and_relegated +
+#     Top_4_total_GD +
+#     Average_goals_by_top_3_players +
+#     Gini_coefficient,
+#   data = CL_scaled_data,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_all_CL_robust)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_all_CL_robust, vcov = vcovHC, type = "HC1")
+
+
+##using principal components
+CL_pca_results <- pca_results_comp_breakdown %>% filter(competition == "CL")
+panel_pc_CL <- plm(
+  European_performance ~
+    PC1 +
+    PC2 +
+    PC3 +
+    PC4 +
+    PC5,
+  data = CL_pca_results,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_pc_CL)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_pc_CL, vcov = vcovHC, type = "HC1")
+
+#rerun with only significant pcs
+# panel_pc_CL_robust <- plm(
+#   European_performance ~
+#     PC1 +
+#     PC3 +
+#     PC4,
+#   data = CL_pca_results,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_pc_CL_robust)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_pc_CL_robust, vcov = vcovHC, type = "HC1")
+
+
+##using competitive indices
+CL_competitive_index <- scaled_full_data_comp_breakdown_competitiveness_index %>% filter(competition == "CL")
+panel_indices_multi_CL <- plm(
+  European_performance ~
+    competitive_title_index +
+    elite_dominance_index +
+    player_rotation_index,
+  data = CL_competitive_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_multi_CL)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_multi_CL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_title_CL <- plm(
+#   European_performance ~ competitive_title_index,
+#   data = CL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_title_CL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_title_CL, vcov = vcovHC, type = "HC1")
+
+panel_indices_dominance_CL <- plm(
+  European_performance ~ elite_dominance_index,
+  data = CL_competitive_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_dominance_CL)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_dominance_CL, vcov = vcovHC, type = "HC1")
+
+panel_indices_competitiveness_CL <- plm(
+  European_performance ~ competitive_balance_index,
+  data = CL_competitive_index,
+  index = c("Country","season_start"),
+  effect = "twoways",
+  model = "within"
+)
+summary(panel_indices_competitiveness_CL)
+#compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+coeftest(panel_indices_competitiveness_CL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_points_CL <- plm(
+#   European_performance ~ multilevel_competitiveness_index,
+#   data = CL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_points_CL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_points_CL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_players_CL <- plm(
+#   European_performance ~ player_rotation_index,
+#   data = CL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_players_CL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_players_CL, vcov = vcovHC, type = "HC1")
+
+
+##panel regression for EL----
+##all data with multicollinearity accounted for
+# EL_scaled_data <- scaled_full_data_comp_breakdown %>% filter(competition == "EL")
+# panel_all_EL <- plm(
+#   European_performance ~
+#     First_and_second +
+#     First_and_CL +
+#     CL_and_relegated +
+#     Top_4_total_GD +
+#     Average_number_of_players +
+#     Average_goals_by_top_3_players +
+#     Gini_coefficient,
+#   data = EL_scaled_data,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_all_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_all_EL, vcov = vcovHC, type = "HC1")
+
+
+##using principal components
+# EL_pca_results <- pca_results_comp_breakdown %>% filter(competition == "EL")
+# panel_pc_EL <- plm(
+#   European_performance ~
+#     PC1 +
+#     PC2 +
+#     PC3 +
+#     PC4 +
+#     PC5,
+#   data = EL_pca_results,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_pc_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_pc_EL, vcov = vcovHC, type = "HC1")
+
+
+##using competitive indices
+# EL_competitive_index <- scaled_full_data_comp_breakdown_competitiveness_index %>% filter(competition == "EL")
+# panel_indices_multi_EL <- plm(
+#   European_performance ~
+#     competitive_title_index +
+#     elite_dominance_index +
+#     player_rotation_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_multi_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_multi_EL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_title_EL <- plm(
+#   European_performance ~ competitive_title_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_title_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_title_EL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_dominance_EL <- plm(
+#   European_performance ~ elite_dominance_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_dominance_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_dominance_EL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_competitiveness_EL <- plm(
+#   European_performance ~ competitive_balance_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_competitiveness_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_competitiveness_EL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_points_EL <- plm(
+#   European_performance ~ multilevel_competitiveness_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_points_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_points_EL, vcov = vcovHC, type = "HC1")
+
+# panel_indices_players_EL <- plm(
+#   European_performance ~ player_rotation_index,
+#   data = EL_competitive_index,
+#   index = c("Country","season_start"),
+#   effect = "twoways",
+#   model = "within"
+# )
+# summary(panel_indices_players_EL)
+# #compute clustered standard errors (adjustment for autocorrelation + heteroskedasticity)
+# coeftest(panel_indices_players_EL, vcov = vcovHC, type = "HC1")
+
+indices_comp_breakdown_plotting_data <- regression_coefficient_data(panel_indices_multi_CL) %>%
+  mutate(term = paste0(term, " (CL multi)")) %>%
+  rbind(regression_coefficient_data(panel_indices_dominance_CL) %>% mutate(term = paste0(term, " (CL)"))) %>%
+  rbind(regression_coefficient_data(panel_indices_competitiveness_CL) %>% mutate(term = paste0(term, " (CL)")))