.Rhistory

##### Setup the environment #####
# Install the required packages (uncomment to install all of them)
#install.packages(c("readr", "RTransferEntropy", "zoo", "e1071", "xts", "doParallel", "foreach", "tidyverse"))
# Load the required packages
library(readr)
library(RTransferEntropy)
library(zoo)
library(e1071)
library(xts)
library(doParallel)
library(foreach)
library(tidyverse)
# Load the data
Adj_Closing_prices <- read_csv("./full_df.csv")
Adj_Closing_prices$Date <- as.Date(Adj_Closing_prices$Date)
##### Point Estimate TE #####
# Get all columns to compare against BTC
all_cols <- colnames(Adj_Closing_prices)
other_cols <- setdiff(all_cols , c("Date", "BTC"))
# Register parallel backend
num_cores <- detectCores() - 1  # Detect number of CPU cores available on the machine and leave one core free for system processes to run
cl <- makeCluster(num_cores) # Get number of cores to be used for parallelization
registerDoParallel(cl) # Initialize the parallel backed for the 'foreach' functions to use the created cluster
# Run parallel foreach loop
results <- foreach(col = other_cols, .combine = rbind, .packages = c("RTransferEntropy", "zoo", "e1071", "xts", "tidyverse")) %dopar% {
cat("Processing", col, "...\n")
temp_df <- Adj_Closing_prices[, c("BTC", col)]
temp_df <- na.omit(temp_df)
if (nrow(temp_df) < 20) {
cat("Skipping", col, "- not enough data after removing NAs.\n")
return(NULL)
}
btc_ret <- diff(log(temp_df$BTC))
stock_ret <- diff(log(temp_df[[col]]))
returns_df <- na.omit(data.frame(btc_ret, stock_ret))
a <- returns_df$btc_ret
b <- returns_df$stock_ret
te <- tryCatch({
transfer_entropy(a, b,
nboot = 1000,
shuffles = 1000,
burn = 0,
seed = 42)
}, error = function(e) {
cat("Failed on", col, ":", e$message, "\n")
return(NULL)
})
if (!is.null(te)) {
coef <- te[3]$coef
data.frame(
Ticker = col,
MI_BTC_to_Ticker = coef[1],
MI_Ticker_to_BTC = coef[[2]],
p_BTC_to_Ticker = coef[7],
p_Ticker_to_BTC = coef[8],
stringsAsFactors = FALSE
)
} else {
NULL
}
}
# Stop cluster
stopCluster(cl)
# View or save results
print(results)
write.csv(results, "20_05_2025_PointEstimate_TransferEntropy_MSTR.csv")
##### Rolling TE #####
# Read the CSV
df <- read.csv("MSTR_BTC.csv", stringsAsFactors = FALSE)
dates <- as.Date(df$Date)
# Ensure numeric and select relevant columns
data <- df %>%
select(BTC, MSTR) %>%
mutate(across(everything(), as.numeric))
rownames(data) <- dates
# Convert to log returns
ret <- diff(log(as.matrix(data)))
ret <- na.omit(ret)
print(ret)
# Convert the data into an xts object
xts_data <- xts(ret, order.by = as.Date(df$Date[-1]))
# Create the list where the first element contains the xts object
data_list <- list(xts_data)
# Rolling Transfer Entropy
window_size <- 252   # ~1 year
step_size <- 1
n_steps <- floor((nrow(data_list[[1]]) - window_size) / step_size)
#time_index <- as.Date(rep(NA, n_steps))
# Prepare output
results_list <- list()
# Set up parallel backend
n_cores <- parallel::detectCores() - 2
cl <- makeCluster(n_cores)
registerDoParallel(cl)
for (ticker_index in 1:length(data_list)) {
stripped_names <- gsub("\\.Adjusted$", "", colnames(data_list[[ticker_index]])[[2]])
print(paste("Processing", stripped_names))
results <- foreach(i = seq(1, nrow(data_list[[ticker_index]]) - window_size+1, by = step_size),
.combine = rbind,
.packages = c("RTransferEntropy", "xts")) %dopar% {
window_x <- data_list[[ticker_index]][,1][i:(i+window_size-1)]
window_y <- data_list[[ticker_index]][,2][i:(i+window_size-1)]
cat("Step", i, ": BTC head =", head(window_x), "MSTR head =", head(window_y), "\n",
file = "te_debug_log.txt", append = TRUE)
if (any(is.na(window_x) | is.na(window_y))) return(rep(NA, 5))
te_xy_result <- suppressMessages(
suppressWarnings(
transfer_entropy(window_x, window_y, lx=1, ly=1,
burn = 0, nboot = 1000,
seed = 42, shuffles = 1000,
na.rm = TRUE)
)
)
as.numeric(c(index(data_list[[ticker_index]])[i + window_size -1],
te_xy_result$coef[1], te_xy_result$coef[7],
te_xy_result$coef[[2]], te_xy_result$coef[8]))
#te_crypto_stock[i] <- te_btc_stock
#te_stock_crypto[i] <- te_stock_btc
#pval_crypto_stock[i] <- p_btc_stock
#pval_stock_crypto[i] <- p_stock_btc
#}
#result_dates <- dates[rolling_indexes[, window_size]]
}
# Convert results to data.frame
df <- as.data.frame(results)
colnames(df) <- c("Date",
paste0("TE BTC-->", stripped_names),
paste0("p_values BTC-->", stripped_names),
paste0("TE ", stripped_names, "-->BTC"),
paste0("p_values ", stripped_names, "-->BTC"))
df$Date <- as.Date(df$Date)
results_list[[stripped_names]] <- df
}
stopCluster(cl)
write.csv(results_list$MSTR, "20_05_2025_Rolling_TransferEntropy_MSTR.csv")