From 17a2a503db34d513c655a7b658dc24e9e1bd0e5c Mon Sep 17 00:00:00 2001
From: iampratik13 <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 15:32:00 +0530
Subject: [PATCH 01/14] create a Time Series Analysis algorithm

---
 quantitative_finance/time_series_analyzer.r | 494 ++++++++++++++++++++
 1 file changed, 494 insertions(+)
 create mode 100644 quantitative_finance/time_series_analyzer.r

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
new file mode 100644
index 00000000..b29b1839
--- /dev/null
+++ b/quantitative_finance/time_series_analyzer.r
@@ -0,0 +1,494 @@
+# Time Series Analysis and ARIMA Modeling
+# Implements comprehensive time series analysis with ARIMA models
+# Features: Stationarity testing, model selection, forecasting, and diagnostics
+
+library(R6)
+
+#' TimeSeriesAnalyzer Class
+#' @description R6 class for time series analysis and ARIMA modeling
+#' @details Provides functionality for:
+#' - Stationarity testing (ADF test)
+#' - ACF/PACF analysis
+#' - ARIMA model fitting
+#' - Model selection using AIC/BIC
+#' - Forecasting with confidence intervals
+TimeSeriesAnalyzer <- R6Class(
+  "TimeSeriesAnalyzer",
+  
+  public = list(
+    #' @description Initialize analyzer with time series data
+    #' @param data Time series data (numeric vector)
+    #' @param frequency Frequency of the time series (default: 1)
+    initialize = function(data = NULL, frequency = 1) {
+      if (!is.null(data)) {
+        private$validate_input(data)
+        self$data <- data
+        self$frequency <- frequency
+        private$n <- length(data)
+      }
+      invisible(self)
+    },
+
+    #' @description Test for stationarity using Augmented Dickey-Fuller test
+    #' @param max_lags Maximum number of lags to consider
+    test_stationarity = function(max_lags = NULL) {
+      if (is.null(self$data)) {
+        stop("No data available. Please initialize with data first.")
+      }
+      
+      if (is.null(max_lags)) {
+        max_lags <- floor(sqrt(private$n))
+      }
+      
+      # Perform ADF test
+      result <- private$adf_test(max_lags)
+      
+      # Store results
+      private$stationarity_results <- result
+      return(result)
+    },
+
+    #' @description Calculate ACF and PACF
+    #' @param max_lag Maximum lag to consider
+    calculate_acf_pacf = function(max_lag = NULL) {
+      if (is.null(self$data)) {
+        stop("No data available. Please initialize with data first.")
+      }
+      
+      if (is.null(max_lag)) {
+        max_lag <- min(private$n - 1, floor(10 * log10(private$n)))
+      }
+      
+      # Calculate ACF
+      acf_result <- private$calculate_acf(max_lag)
+      
+      # Calculate PACF
+      pacf_result <- private$calculate_pacf(max_lag)
+      
+      return(list(
+        acf = acf_result,
+        pacf = pacf_result,
+        lags = 1:max_lag
+      ))
+    },
+
+    #' @description Fit ARIMA model to the data
+    #' @param p AR order
+    #' @param d Differencing order
+    #' @param q MA order
+    fit_arima = function(p = 1, d = 0, q = 1) {
+      if (is.null(self$data)) {
+        stop("No data available. Please initialize with data first.")
+      }
+      
+      # Validate parameters
+      if (any(c(p, d, q) < 0)) {
+        stop("ARIMA orders must be non-negative")
+      }
+      
+      # Fit ARIMA model
+      model <- private$fit_arima_model(p, d, q)
+      private$current_model <- model
+      
+      return(model)
+    },
+
+    #' @description Automatic model selection using AIC
+    #' @param max_p Maximum AR order to consider
+    #' @param max_d Maximum differencing order
+    #' @param max_q Maximum MA order
+    select_best_model = function(max_p = 3, max_d = 2, max_q = 3) {
+      if (is.null(self$data)) {
+        stop("No data available. Please initialize with data first.")
+      }
+      
+      best_aic <- Inf
+      best_order <- c(0, 0, 0)
+      
+      # Grid search over possible orders
+      for (p in 0:max_p) {
+        for (d in 0:max_d) {
+          for (q in 0:max_q) {
+            tryCatch({
+              model <- self$fit_arima(p, d, q)
+              if (model$aic < best_aic) {
+                best_aic <- model$aic
+                best_order <- c(p, d, q)
+                private$best_model <- model
+              }
+            }, error = function(e) {
+              # Skip failed models
+            })
+          }
+        }
+      }
+      
+      return(list(
+        order = best_order,
+        aic = best_aic
+      ))
+    },
+
+    #' @description Generate forecasts with confidence intervals
+    #' @param h Forecast horizon
+    #' @param level Confidence level (0-1)
+    forecast = function(h = 10, level = 0.95) {
+      if (is.null(private$current_model)) {
+        stop("No model fitted. Please fit a model first.")
+      }
+      
+      # Generate forecasts
+      forecasts <- private$generate_forecasts(h, level)
+      return(forecasts)
+    },
+
+    #' @description Perform model diagnostics
+    diagnose_model = function() {
+      if (is.null(private$current_model)) {
+        stop("No model fitted. Please fit a model first.")
+      }
+      
+      residuals <- private$current_model$residuals
+      
+      # Calculate diagnostic statistics
+      diagnostics <- list(
+        residual_mean = mean(residuals),
+        residual_sd = sd(residuals),
+        ljung_box = private$ljung_box_test(residuals),
+        normality = private$normality_test(residuals),
+        arch_effect = private$arch_test(residuals)
+      )
+      
+      return(diagnostics)
+    },
+
+    # Public fields
+    data = NULL,
+    frequency = NULL
+  ),
+  
+  private = list(
+    n = NULL,
+    current_model = NULL,
+    best_model = NULL,
+    stationarity_results = NULL,
+    
+    validate_input = function(data) {
+      if (!is.numeric(data)) {
+        stop("Input data must be numeric")
+      }
+      if (any(is.na(data))) {
+        stop("Input data contains missing values")
+      }
+      if (length(data) < 3) {
+        stop("Input data must have at least 3 observations")
+      }
+    },
+    
+    adf_test = function(max_lags) {
+      y <- self$data
+      n <- length(y)
+      
+      # Calculate first differences
+      dy <- diff(y)
+      y_1 <- y[-n]
+      
+      # Construct regression matrix
+      X <- matrix(1, n-1, 1)
+      X <- cbind(X, y_1)
+      
+      # Add lagged differences
+      for (lag in 1:max_lags) {
+        X <- cbind(X, stats::lag(dy, lag)[-c(1:lag)])
+      }
+      
+      # Remove NA rows
+      complete_cases <- stats::complete.cases(X)
+      X <- X[complete_cases, ]
+      dy <- dy[complete_cases]
+      
+      # Fit regression
+      fit <- stats::lm(dy ~ X - 1)
+      
+      # Calculate test statistic
+      coef <- stats::coef(fit)[2]
+      se <- sqrt(diag(stats::vcov(fit)))[2]
+      t_stat <- (coef - 1) / se
+      
+      # Critical values (approximate)
+      crit_values <- c(
+        "1%" = -3.43,
+        "5%" = -2.86,
+        "10%" = -2.57
+      )
+      
+      return(list(
+        statistic = t_stat,
+        critical_values = crit_values,
+        is_stationary = t_stat < -2.86  # 5% level
+      ))
+    },
+    
+    calculate_acf = function(max_lag) {
+      y <- scale(self$data)  # Standardize data
+      n <- length(y)
+      acf <- numeric(max_lag)
+      
+      for (k in 1:max_lag) {
+        acf[k] <- stats::cor(y[1:(n-k)], y[(k+1):n])
+      }
+      
+      return(acf)
+    },
+    
+    calculate_pacf = function(max_lag) {
+      acf <- private$calculate_acf(max_lag)
+      pacf <- numeric(max_lag)
+      
+      # Durbin-Levinson algorithm
+      for (k in 1:max_lag) {
+        if (k == 1) {
+          pacf[k] <- acf[1]
+        } else {
+          r <- matrix(0, k, k)
+          for (i in 1:k) {
+            for (j in 1:k) {
+              r[i,j] <- acf[abs(i-j)+1]
+            }
+          }
+          b <- acf[1:k]
+          pacf[k] <- solve(r, b)[k]
+        }
+      }
+      
+      return(pacf)
+    },
+    
+    fit_arima_model = function(p, d, q) {
+      # Difference data
+      y <- self$data
+      for (i in 1:d) {
+        y <- diff(y)
+      }
+      
+      # Construct and solve Yule-Walker equations for AR part
+      if (p > 0) {
+        r <- private$calculate_acf(p)
+        phi <- solve(stats::toeplitz(r[1:p]), r[2:(p+1)])
+      } else {
+        phi <- numeric(0)
+      }
+      
+      # Estimate MA parameters using innovation algorithm
+      if (q > 0) {
+        theta <- numeric(q)
+        e <- y
+        for (i in 1:10) {  # Iterate to improve estimates
+          r <- private$calculate_acf(q)
+          psi <- numeric(q)
+          for (j in 1:q) {
+            psi[j] <- sum(theta[1:(j-1)] * rev(psi[1:(j-1)])) + theta[j]
+          }
+          theta <- solve(stats::toeplitz(c(1, psi[1:(q-1)])), r[1:q])
+        }
+      } else {
+        theta <- numeric(0)
+      }
+      
+      # Calculate residuals and AIC
+      resid <- private$calculate_residuals(y, phi, theta)
+      n_params <- p + q
+      aic <- length(resid) * log(var(resid)) + 2 * n_params
+      
+      return(list(
+        coefficients = list(ar = phi, ma = theta),
+        residuals = resid,
+        aic = aic,
+        order = c(p, d, q)
+      ))
+    },
+    
+    calculate_residuals = function(y, phi, theta) {
+      n <- length(y)
+      p <- length(phi)
+      q <- length(theta)
+      resid <- numeric(n)
+      
+      for (t in (max(p,q)+1):n) {
+        pred <- 0
+        if (p > 0) {
+          pred <- pred + sum(phi * y[(t-1):(t-p)])
+        }
+        if (q > 0) {
+          pred <- pred + sum(theta * resid[(t-1):(t-q)])
+        }
+        resid[t] <- y[t] - pred
+      }
+      
+      return(resid[(max(p,q)+1):n])
+    },
+    
+    generate_forecasts = function(h, level) {
+      model <- private$current_model
+      y <- self$data
+      n <- length(y)
+      
+      # Get model orders
+      p <- length(model$coefficients$ar)
+      d <- model$order[2]
+      q <- length(model$coefficients$ma)
+      
+      # Generate point forecasts
+      forecasts <- numeric(h)
+      for (i in 1:h) {
+        pred <- 0
+        if (p > 0) {
+          ar_terms <- if (i <= p) {
+            y[(n-p+i):n]
+          } else {
+            forecasts[(i-p):(i-1)]
+          }
+          pred <- pred + sum(model$coefficients$ar * ar_terms)
+        }
+        forecasts[i] <- pred
+      }
+      
+      # Calculate prediction intervals
+      sigma <- sd(model$residuals)
+      z <- stats::qnorm((1 + level) / 2)
+      se <- sigma * sqrt(cumsum(rep(1, h)))
+      lower <- forecasts - z * se
+      upper <- forecasts + z * se
+      
+      return(list(
+        mean = forecasts,
+        lower = lower,
+        upper = upper,
+        level = level
+      ))
+    },
+    
+    ljung_box_test = function(residuals) {
+      max_lag <- min(20, length(residuals) - 1)
+      acf <- private$calculate_acf(max_lag)
+      n <- length(residuals)
+      
+      Q <- n * (n + 2) * sum((acf^2) / (n - 1:max_lag))
+      p_value <- 1 - stats::pchisq(Q, max_lag)
+      
+      return(list(
+        statistic = Q,
+        p_value = p_value
+      ))
+    },
+    
+    normality_test = function(residuals) {
+      # Jarque-Bera test
+      n <- length(residuals)
+      s <- sum((residuals - mean(residuals))^3) / (n * sd(residuals)^3)  # skewness
+      k <- sum((residuals - mean(residuals))^4) / (n * sd(residuals)^4) - 3  # excess kurtosis
+      JB <- n * (s^2/6 + k^2/24)
+      p_value <- 1 - stats::pchisq(JB, df = 2)
+      
+      return(list(
+        statistic = JB,
+        p_value = p_value
+      ))
+    },
+    
+    arch_test = function(residuals) {
+      # ARCH LM test
+      sq_resid <- residuals^2
+      n <- length(sq_resid)
+      lags <- 5
+      
+      # Regression of squared residuals on lagged squared residuals
+      X <- matrix(1, n-lags, 1)
+      for (i in 1:lags) {
+        X <- cbind(X, stats::lag(sq_resid, i)[-c(1:lags)])
+      }
+      
+      fit <- stats::lm(sq_resid[-c(1:lags)] ~ X - 1)
+      R2 <- summary(fit)$r.squared
+      LM <- n * R2
+      p_value <- 1 - stats::pchisq(LM, lags)
+      
+      return(list(
+        statistic = LM,
+        p_value = p_value
+      ))
+    }
+  )
+)
+
+# Demonstration
+demonstrate_time_series_analysis <- function() {
+  cat("=== Time Series Analysis Demo ===\n\n")
+  
+  # Generate sample time series
+  set.seed(42)
+  n <- 500
+  
+  # AR(1) process with trend and seasonality
+  t <- 1:n
+  trend <- 0.01 * t
+  seasonal <- 2 * sin(2 * pi * t / 12)
+  ar_process <- stats::arima.sim(list(ar = 0.7), n = n)
+  y <- trend + seasonal + ar_process
+  
+  # Initialize analyzer
+  ts_analyzer <- TimeSeriesAnalyzer$new(y, frequency = 12)
+  
+  # Test stationarity
+  cat("Testing for stationarity...\n")
+  stat_test <- ts_analyzer$test_stationarity()
+  cat(sprintf("ADF test statistic: %.3f\n", stat_test$statistic))
+  cat("Critical values:\n")
+  print(stat_test$critical_values)
+  cat(sprintf("Series is %sstationary at 5%% level\n\n",
+              ifelse(stat_test$is_stationary, "", "non-")))
+  
+  # Analyze ACF/PACF
+  cat("Calculating ACF and PACF...\n")
+  corr <- ts_analyzer$calculate_acf_pacf(20)
+  cat("First 5 lags:\n")
+  cat("ACF:  ")
+  cat(sprintf("%.3f ", corr$acf[1:5]))
+  cat("\nPACF: ")
+  cat(sprintf("%.3f ", corr$pacf[1:5]))
+  cat("\n\n")
+  
+  # Select best model
+  cat("Selecting best ARIMA model...\n")
+  best <- ts_analyzer$select_best_model(max_p = 2, max_d = 1, max_q = 2)
+  cat(sprintf("Best model: ARIMA(%d,%d,%d)\n", 
+              best$order[1], best$order[2], best$order[3]))
+  cat(sprintf("AIC: %.2f\n\n", best$aic))
+  
+  # Generate forecasts
+  cat("Generating forecasts...\n")
+  h <- 12  # Forecast horizon
+  forecasts <- ts_analyzer$forecast(h = h, level = 0.95)
+  
+  cat("Point forecasts for next 12 periods:\n")
+  cat(sprintf("%.2f ", forecasts$mean))
+  cat("\n\n")
+  
+  # Model diagnostics
+  cat("Performing model diagnostics...\n")
+  diagnostics <- ts_analyzer$diagnose_model()
+  
+  cat("Residual diagnostics:\n")
+  cat(sprintf("Mean: %.3f\n", diagnostics$residual_mean))
+  cat(sprintf("Standard deviation: %.3f\n", diagnostics$residual_sd))
+  cat(sprintf("Ljung-Box test p-value: %.3f\n", diagnostics$ljung_box$p_value))
+  cat(sprintf("Normality test p-value: %.3f\n", diagnostics$normality$p_value))
+  cat(sprintf("ARCH test p-value: %.3f\n", diagnostics$arch_effect$p_value))
+  
+  cat("\n=== Demo Complete ===\n")
+}
+
+# Run demonstration if not in interactive mode
+if (!interactive()) {
+  demonstrate_time_series_analysis()
+}
\ No newline at end of file

From 4b4844e3c8fdf53711b3531f8d095ee13f1a6741 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:27:28 +0530
Subject: [PATCH 02/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index b29b1839..7f769478 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -253,7 +253,12 @@ TimeSeriesAnalyzer <- R6Class(
           r <- matrix(0, k, k)
           for (i in 1:k) {
             for (j in 1:k) {
-              r[i,j] <- acf[abs(i-j)+1]
+              idx <- abs(i-j)+1
+              if (idx <= length(acf)) {
+                r[i,j] <- acf[idx]
+              } else {
+                r[i,j] <- 0
+              }
             }
           }
           b <- acf[1:k]

From 0c6a999deeabf607c618d7987bd365d8ddada8fe Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:27:40 +0530
Subject: [PATCH 03/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 7f769478..4e8bc62f 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -278,7 +278,7 @@ TimeSeriesAnalyzer <- R6Class(
       
       # Construct and solve Yule-Walker equations for AR part
       if (p > 0) {
-        r <- private$calculate_acf(p)
+        r <- private$calculate_acf(p + 1)
         phi <- solve(stats::toeplitz(r[1:p]), r[2:(p+1)])
       } else {
         phi <- numeric(0)

From 678400a4ebef1711dc6a069fb940cc44280e3843 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:27:47 +0530
Subject: [PATCH 04/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 4e8bc62f..19a2bf77 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -294,7 +294,11 @@ TimeSeriesAnalyzer <- R6Class(
           for (j in 1:q) {
             psi[j] <- sum(theta[1:(j-1)] * rev(psi[1:(j-1)])) + theta[j]
           }
-          theta <- solve(stats::toeplitz(c(1, psi[1:(q-1)])), r[1:q])
+          if (q == 1) {
+            theta <- solve(stats::toeplitz(1), r[1])
+          } else {
+            theta <- solve(stats::toeplitz(c(1, psi[1:(q-1)])), r[1:q])
+          }
         }
       } else {
         theta <- numeric(0)

From 809b9027d489cd0eee9defd2f99b69cff4de5a00 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:27:57 +0530
Subject: [PATCH 05/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 19a2bf77..1ff8e620 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -199,7 +199,8 @@ TimeSeriesAnalyzer <- R6Class(
       
       # Add lagged differences
       for (lag in 1:max_lags) {
-        X <- cbind(X, stats::lag(dy, lag)[-c(1:lag)])
+        # Use backward lag: pad with NA, then take first (length-dy-lag) elements
+        X <- cbind(X, c(rep(NA, lag), dy[1:(length(dy)-lag)])[-c(1:lag)])
       }
       
       # Remove NA rows

From f927a9fc9cc66f763f820e56933b81038b8d8e36 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:28:10 +0530
Subject: [PATCH 06/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 1ff8e620..8a9e3449 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -412,15 +412,17 @@ TimeSeriesAnalyzer <- R6Class(
       n <- length(sq_resid)
       lags <- 5
       
-      # Regression of squared residuals on lagged squared residuals
-      X <- matrix(1, n-lags, 1)
-      for (i in 1:lags) {
-        X <- cbind(X, stats::lag(sq_resid, i)[-c(1:lags)])
+      # Use embed to construct lagged matrix (current and past lags)
+      # embed(sq_resid, lags + 1) returns a matrix with columns: t, t-1, ..., t-lags
+      if (n <= lags) {
+        stop("Not enough observations for ARCH test lags.")
       }
-      
-      fit <- stats::lm(sq_resid[-c(1:lags)] ~ X - 1)
+      lagged_mat <- embed(sq_resid, lags + 1)
+      y <- lagged_mat[, 1]  # current squared residuals
+      X <- lagged_mat[, -1] # lagged squared residuals
+      fit <- stats::lm(y ~ X)
       R2 <- summary(fit)$r.squared
-      LM <- n * R2
+      LM <- nrow(lagged_mat) * R2
       p_value <- 1 - stats::pchisq(LM, lags)
       
       return(list(

From 4f57343ca2f74ef9a7016953eed5457ae53c3c17 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:28:20 +0530
Subject: [PATCH 07/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 8a9e3449..3433b4cf 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -380,10 +380,10 @@ TimeSeriesAnalyzer <- R6Class(
     
     ljung_box_test = function(residuals) {
       max_lag <- min(20, length(residuals) - 1)
-      acf <- private$calculate_acf(max_lag)
+      acf_vals <- acf(residuals, plot = FALSE, lag.max = max_lag)$acf[-1]
       n <- length(residuals)
       
-      Q <- n * (n + 2) * sum((acf^2) / (n - 1:max_lag))
+      Q <- n * (n + 2) * sum((acf_vals^2) / (n - 1:max_lag))
       p_value <- 1 - stats::pchisq(Q, max_lag)
       
       return(list(

From 85af7fc7f9a4a43cf395f2638615b3f1a7efb821 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:52:11 +0530
Subject: [PATCH 08/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 3433b4cf..b206c775 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -198,9 +198,12 @@ TimeSeriesAnalyzer <- R6Class(
       X <- cbind(X, y_1)
       
       # Add lagged differences
-      for (lag in 1:max_lags) {
-        # Use backward lag: pad with NA, then take first (length-dy-lag) elements
-        X <- cbind(X, c(rep(NA, lag), dy[1:(length(dy)-lag)])[-c(1:lag)])
+      if (max_lags > 0) {
+        # Add lagged differences using embed() for clarity
+        lagged_dy <- embed(dy, max_lags + 1)[, -1, drop = FALSE]
+        X <- X[(max_lags+1):nrow(X), ]  # Align X with lagged_dy rows
+        X <- cbind(X, lagged_dy)
+        dy <- dy[(max_lags+1):length(dy)]
       }
       
       # Remove NA rows

From 75448f341dc17884fbaca262de60d45709da337b Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:52:25 +0530
Subject: [PATCH 09/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index b206c775..72885bfa 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -217,7 +217,7 @@ TimeSeriesAnalyzer <- R6Class(
       # Calculate test statistic
       coef <- stats::coef(fit)[2]
       se <- sqrt(diag(stats::vcov(fit)))[2]
-      t_stat <- (coef - 1) / se
+      t_stat <- coef / se
       
       # Critical values (approximate)
       crit_values <- c(

From ad72bf6014b3a2d870f7e8a2bd5723a5ec004d78 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:52:33 +0530
Subject: [PATCH 10/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 72885bfa..0545788a 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -383,7 +383,7 @@ TimeSeriesAnalyzer <- R6Class(
     
     ljung_box_test = function(residuals) {
       max_lag <- min(20, length(residuals) - 1)
-      acf_vals <- acf(residuals, plot = FALSE, lag.max = max_lag)$acf[-1]
+      acf_vals <- stats::acf(residuals, plot = FALSE, lag.max = max_lag)$acf[-1]
       n <- length(residuals)
       
       Q <- n * (n + 2) * sum((acf_vals^2) / (n - 1:max_lag))

From 4b97f5b598c7e9c6591a26abdaea34e5ccb4f2ed Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:52:46 +0530
Subject: [PATCH 11/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 0545788a..883824ae 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -420,7 +420,7 @@ TimeSeriesAnalyzer <- R6Class(
       if (n <= lags) {
         stop("Not enough observations for ARCH test lags.")
       }
-      lagged_mat <- embed(sq_resid, lags + 1)
+      lagged_mat <- stats::embed(sq_resid, lags + 1)
       y <- lagged_mat[, 1]  # current squared residuals
       X <- lagged_mat[, -1] # lagged squared residuals
       fit <- stats::lm(y ~ X)

From afb68ec5d9e3cbb60a125961dad5ee66a5fc2acf Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:52:58 +0530
Subject: [PATCH 12/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 883824ae..cc4a3861 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -357,7 +357,7 @@ TimeSeriesAnalyzer <- R6Class(
         pred <- 0
         if (p > 0) {
           ar_terms <- if (i <= p) {
-            y[(n-p+i):n]
+            y[(n-p+1):n]
           } else {
             forecasts[(i-p):(i-1)]
           }

From f3bb26c0718572449df523f36816e28a02588dc8 Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:56:42 +0530
Subject: [PATCH 13/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index cc4a3861..6c0d0a8c 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -356,10 +356,21 @@ TimeSeriesAnalyzer <- R6Class(
       for (i in 1:h) {
         pred <- 0
         if (p > 0) {
-          ar_terms <- if (i <= p) {
-            y[(n-p+1):n]
+          # Use the most recent p values (from original data and previous forecasts)
+          if (i <= p) {
+            ar_terms <- y[(n-p+i):(n+i-1)]
           } else {
-            forecasts[(i-p):(i-1)]
+            # Combine tail of y and head of forecasts as needed
+            num_from_y <- max(0, p - (i-1))
+            num_from_forecasts <- p - num_from_y
+            if (num_from_y > 0) {
+              ar_terms <- c(
+                y[(n - p + i):(n)],
+                forecasts[1:num_from_forecasts]
+              )
+            } else {
+              ar_terms <- forecasts[(i-p):(i-1)]
+            }
           }
           pred <- pred + sum(model$coefficients$ar * ar_terms)
         }

From 4e11f605aca9e44b903948bfd221de7286bf30fc Mon Sep 17 00:00:00 2001
From: Pratik <bhagwatpratik781@gmail.com>
Date: Sun, 12 Oct 2025 18:56:53 +0530
Subject: [PATCH 14/14] Update quantitative_finance/time_series_analyzer.r

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 quantitative_finance/time_series_analyzer.r | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/quantitative_finance/time_series_analyzer.r b/quantitative_finance/time_series_analyzer.r
index 6c0d0a8c..96a5e01c 100644
--- a/quantitative_finance/time_series_analyzer.r
+++ b/quantitative_finance/time_series_analyzer.r
@@ -296,7 +296,11 @@ TimeSeriesAnalyzer <- R6Class(
           r <- private$calculate_acf(q)
           psi <- numeric(q)
           for (j in 1:q) {
-            psi[j] <- sum(theta[1:(j-1)] * rev(psi[1:(j-1)])) + theta[j]
+            if (j > 1) {
+              psi[j] <- sum(theta[1:(j-1)] * rev(psi[1:(j-1)])) + theta[j]
+            } else {
+              psi[j] <- theta[j]
+            }
           }
           if (q == 1) {
             theta <- solve(stats::toeplitz(1), r[1])