The R package spectralBacktest implements a class of backtests introduced by Michael Gordy and Alexander McNeil in the paper ``Spectral backtests of forecast distributions with application to risk management'' (FEDS 2018-021).

The package provides code for all the kernel functions implemented in our empirical exercises, including the multinomial Pearson kernel, the beta kernel, and the truncated probitnormal score kernel. For the tests of conditional coverage, the package provides the four conditioning variable transformations (CVT) that were used.

The package is designed with two objectives in mind:

  1. Uniform syntax. Uniformity allows the user to apply() a list of tests to a given dataset with a single statement. This is particularly useful for simulation studies of test size and power.
  2. Extensibility. As demonstrated in the vignettes, it is straightforward and quick to write custom kernel functions and CVT.

Download version 0.1.2 (39K).

Michael B Gordy
Last update: 26-Mar-2018