## ----setup, include = FALSE--------------------------------------------------- knitr::opts_chunk$set( collapse = TRUE, comment = "#>", eval = FALSE ) ## ----basic-------------------------------------------------------------------- # library(EventStudy) # # # Size analysis: no abnormal return injected # size_result <- simulate_event_study( # n_events = 20, # event_window = c(-5, 5), # estimation_window_length = 120, # abnormal_return = 0, # H0 is true # n_simulations = 1000, # seed = 42 # ) # print(size_result) # #> Event Study Simulation # #> N events: 20 # #> Event window: [ -5 , 5 ] # #> Abnormal return: 0 # #> Test statistic: CSectT # #> Alpha: 0.05 # #> N simulations: 1000 # #> Power (day 0): 0.048 ## ----power-------------------------------------------------------------------- # # Power analysis: 2% abnormal return on event day # power_result <- simulate_event_study( # n_events = 20, # event_window = c(-5, 5), # estimation_window_length = 120, # abnormal_return = 0.02, # 2% effect # n_simulations = 1000, # seed = 42 # ) # print(power_result) # #> Power (day 0): 0.723 ## ----effect-sizes------------------------------------------------------------- # effects <- c(0, 0.005, 0.01, 0.02, 0.05) # power_curve <- sapply(effects, function(ar) { # res <- simulate_event_study( # n_events = 20, # abnormal_return = ar, # n_simulations = 500, # seed = 42 # ) # res$power # }) # # plot(effects, power_curve, type = "b", # xlab = "Abnormal Return", ylab = "Power", # main = "Power Curve") # abline(h = 0.05, lty = 2, col = "grey") ## ----sample-sizes------------------------------------------------------------- # sample_sizes <- c(5, 10, 20, 50, 100) # power_by_n <- sapply(sample_sizes, function(n) { # res <- simulate_event_study( # n_events = n, # abnormal_return = 0.01, # n_simulations = 500, # seed = 42 # ) # res$power # }) # # plot(sample_sizes, power_by_n, type = "b", # xlab = "Number of Events", ylab = "Power", # main = "Power vs Sample Size (AR = 1%)") ## ----dgp---------------------------------------------------------------------- # result <- simulate_event_study( # n_events = 20, # abnormal_return = 0.01, # n_simulations = 500, # dgp_params = list( # alpha = 0.0001, # drift # beta = 1.2, # market exposure # sigma_firm = 0.015, # idiosyncratic volatility # sigma_market = 0.02 # market volatility # ), # seed = 42 # ) ## ----test-stats--------------------------------------------------------------- # tests <- c("CSectT", "PatellZ", "BMP", "SignT", "RankT") # power_by_test <- sapply(tests, function(ts) { # res <- simulate_event_study( # n_events = 20, # abnormal_return = 0.01, # test_statistic = ts, # n_simulations = 500, # seed = 42 # ) # res$power # }) # # barplot(power_by_test, main = "Power by Test Statistic (AR = 1%)", # ylab = "Power", las = 2) # abline(h = 0.05, lty = 2, col = "grey") ## ----rejection-by-day--------------------------------------------------------- # result <- simulate_event_study( # n_events = 20, # abnormal_return = 0.02, # n_simulations = 500, # seed = 42 # ) # # result$rejection_by_day # #> # A tibble: 11 x 2 # #> relative_index rejection_rate # #> # #> 1 -5 0.048 # #> 2 -4 0.052 # #> ... # #> 6 0 0.723 <-- event day # #> ... ## ----seed--------------------------------------------------------------------- # r1 <- simulate_event_study(n_events = 10, n_simulations = 100, seed = 42) # r2 <- simulate_event_study(n_events = 10, n_simulations = 100, seed = 42) # identical(r1$power, r2$power) # TRUE .