data_load_analysis.R

#' main load analysis data
#'
#'
#'
#' @param portfolio_data portfolio_data
#' @param portfolio_crispy_merge_cols portfolio_crispy_merge_cols
#' @param multi_crispy_data multi_crispy_data
#'
#'
#' @export
main_load_analysis_data <-
  function(multi_crispy_data,
           portfolio_data,
           portfolio_crispy_merge_cols) {

    stopifnot(unique(portfolio_data$asset_type) %in% c("equity", "fixed_income"))

    analysis_data <-
      create_analysis_data(portfolio_data, multi_crispy_data, portfolio_crispy_merge_cols) |>
        aggregate_equities() |>
        compute_analysis_metrics() 

    return(analysis_data)
  }


#' Title
#'
#' @param portfolio_data portfolio_data
#' @param multi_crispy_data multi_crispy_data
#' @param portfolio_crispy_merge_cols portfolio_crispy_merge_cols
#'
create_analysis_data <-
  function(portfolio_data,
           multi_crispy_data,
           portfolio_crispy_merge_cols) {
    # If no portfolio is provided, fill the merging columns
    # with the ones available in crispy, in order to get a full crispy output
    if (nrow(portfolio_data) == 0) {
      merge_cols_values <- multi_crispy_data |>
        dplyr::distinct_at(portfolio_crispy_merge_cols)
      portfolio_data <- dplyr::full_join(portfolio_data, merge_cols_values)
    }

    analysis_data <-
      dplyr::inner_join(
        portfolio_data,
        multi_crispy_data,
        by = portfolio_crispy_merge_cols,
        relationship = "many-to-many"
      )

    facts <- c(
      "net_present_value_baseline",  "net_present_value_shock", "pd_baseline", "pd_shock",
      "exposure_value_usd", "loss_given_default", "pd_portfolio")


    stopifnot(all(facts %in% names(analysis_data))) 

    return(analysis_data)
  }

#' Compute Analysis Metrics
#'
#' @description Function computing financial metrics to use for analysis
#' @param analysis_data analysis_data
#'
compute_analysis_metrics <- function(analysis_data) {
  analysis_data <- analysis_data |>
    dplyr::mutate(
      net_present_value_difference = .data$net_present_value_shock - .data$net_present_value_baseline,
      crispy_perc_value_change = .data$net_present_value_difference / .data$net_present_value_baseline,
      crispy_value_loss = .data$crispy_perc_value_change * .data$exposure_value_usd,
      exposure_at_default = .data$exposure_value_usd * .data$loss_given_default,
      # exposure_at_default_baseline = .data$net_present_value_baseline * .data$loss_given_default,
      # exposure_at_default_shock = .data$net_present_value_shock * .data$loss_given_default,

      # pd_difference_portfolio = .data$pd_portfolio - .data$pd_shock,
      pd_difference = .data$pd_shock - .data$pd_baseline,
      # crispy_perc_pd_change = .data$pd_difference / .data$pd_baseline,

      expected_loss_portfolio = - .data$exposure_at_default * .data$pd_portfolio,
      expected_loss_baseline = - .data$exposure_at_default * .data$pd_baseline,
      expected_loss_shock = - .data$exposure_at_default * .data$pd_shock,
      expected_loss_difference = - .data$exposure_at_default * .data$pd_difference
    )



  return(analysis_data)
}


#' 
#'
#' @description This will do the average of all values of equities over the terms (ie the term disappears and value was just dupicated)
#' @param analysis_data analysis_data
#'
aggregate_equities <- function(analysis_data) {
    facts <- c(
      "net_present_value_baseline",  "net_present_value_shock", "pd_baseline", "pd_shock",
      "exposure_value_usd", "loss_given_default", "pd_portfolio")

    equities_subgroup_analysis <- analysis_data |> 
            dplyr::filter(asset_type=="equity") |>
            dplyr::group_by_at(colnames(analysis_data)[!colnames(analysis_data) %in% c(facts, "term")]) |>
            dplyr::summarise(
              exposure_value_usd=stats::median(.data$exposure_value_usd, na.rm=T),
              net_present_value_baseline=stats::median(.data$net_present_value_baseline, na.rm=T),
              net_present_value_shock=stats::median(.data$net_present_value_shock, na.rm=T),
              pd_baseline=stats::median(.data$pd_baseline, na.rm=T),
              pd_shock=stats::median(.data$pd_shock, na.rm=T),
              loss_given_default=stats::median(.data$loss_given_default, na.rm=T),
              pd_portfolio=stats::median(.data$pd_portfolio, na.rm=T)
            ) |>
            dplyr::ungroup() |>
            dplyr::mutate(
              term=NA,
              loss_given_default=NA
            )
    non_equities_subgroup_analysis  <- analysis_data |>
    dplyr::filter((asset_type != "equity") | is.na(asset_type))

    analysis_data <- dplyr::bind_rows(equities_subgroup_analysis,non_equities_subgroup_analysis)
    return(analysis_data)
}