Skip to contents

Compute rolling annualized tracking difference statistics

Source: R/fund_calc.R
roll_diffs.Rd

For each fund–index pair in fund_index_map, computes rolling, annualized tracking differences over a backward-looking window of n_days calendar days. Both log-return and CAGR forms are returned.

Usage

roll_diffs(
  df,
  n_days,
  fund_index_map,
  date_col = "date",
  index_level = c("net", "gross"),
  annual_days = 365,
  messages = c("roll", "skip"),
  gross_suffix = "-GR"
)

Arguments

df

Data frame containing a date column, fund columns, and benchmark/index columns referenced in fund_index_map.

n_days

Rolling lookback window in calendar days.

fund_index_map

Named character vector mapping fund column names to their corresponding benchmark/index base column names.

date_col

Name of the date column in df. Must be of class Date and sorted (strictly increasing).

index_level

Which index level to use, one of "net" or "gross". If "gross", gross_suffix is appended to the mapped index base name before lookup in df.

annual_days

Number of days used for annualization.

messages

Character vector controlling emitted messages. Any of "roll" (per-pair progress) and "skip" (skip reasons). Use messages = "roll" to show only progress, messages = "skip" to show only skip reasons, or messages = character() or NULL to silence all messages.

gross_suffix

Suffix appended to the mapped index base name when index_level = "gross".

Value

A named list with two data frames, cagr and log. Each data frame contains date_col followed by one column per fund (named as in fund_index_map), holding the rolling annualized tracking differences.

Details

For each date \(t\), a target anchor threshold \(t - n\_days\) is formed. The anchor date \(t_0\) is chosen as the last available observation at or before \(t - n\_days\) among rows where both fund and index values are present. Let \(\Delta = t - t_0\) in calendar days (\(\Delta\) can be greater than n_days when data are missing around the threshold).

The annualized tracking differences are:

  • Log-return difference: $$ \left[\ln\left(\frac{f_t}{f_{t_0}}\right) - \ln\left(\frac{i_t}{i_{t_0}}\right)\right] \times \frac{annual\_days}{\Delta} $$

  • CAGR difference: $$ \left(\frac{f_t}{f_{t_0}}\right)^{annual\_days/\Delta} - \left(\frac{i_t}{i_{t_0}}\right)^{annual\_days/\Delta} $$

#' Values are NA when an anchor cannot be found, current-date inputs are missing, or inputs are invalid for the chosen formula (e.g. any non-positive level for log returns, or non-finite / non-positive ratios for CAGR).

Funds are skipped (optionally with a message) when the fund column is missing, the mapped index column is missing (after applying index_level / gross_suffix), or when fund == index (self-tracking).

Emitted messages will be visible at verbosity level >= 1 (option fundsr.verbosity). Verbosity level >= 4 forces both message types regardless of the messages argument.

See also

Other rolling difference functions: plot_roll_diffs()