fpa_evaluate_values: Evaluate recovered values or an estimated density against the...

In qdfauction: Nonparametric Quantile Density Estimation and Structural Analysis of First-Price Auctions

Evaluate recovered values or an estimated density against the truth

Description

Performance criteria used in the papers behind this package.

'fpa_evaluate_values()' scores recovered private values (Doosti, Dewan and Talebian 2025): 'mise', the mean integrated squared error between a Gaussian kernel density (Sheather–Jones bandwidth) of the values and the true density; 'loglik', \sum_i \log f(\hat v_i); 'ks', the Kolmogorov–Smirnov distance between the empirical distribution of the values and F; and 'msep', the mean squared error against the true values when they are supplied (Doosti 2026a), with 'msep_trim' excluding a share 'trim' at each end of the true-value distribution.

'fpa_evaluate_density()' scores an [fpa_density()] object on its grid: 'mise' over the whole grid and over the upper 'tail' share of the support, and the expected log-likelihood \int f \log \hat f (Doosti, Dewan and Sbai 2026).

Usage

fpa_evaluate_values(values, pdf, cdf = NULL, true = NULL, trim = 0.05)

fpa_evaluate_density(density, pdf, tail = 0.1)

Arguments

values	numeric vector of recovered values, or an [fpa_values()] object.
pdf, cdf	true density and distribution functions.
true	optional true values, aligned with 'values' (for an [fpa_values()] object, in the layout of the original bids).
trim	trimming share for 'msep_trim'.
density	an [fpa_density()] object.
tail	share of the upper support for the tail MISE.

Value

a named numeric vector.

Examples

set.seed(1)
sim <- fpa_simulate(20, 5, Q = function(u) qgamma(u, 10, 2))
fv  <- fpa_values(sim$bids, method = "bernstein", bandwidth = "bcv", monotone = "spline")
fpa_evaluate_values(fv, pdf = function(x) dgamma(x, 10, 2),
                    cdf = function(x) pgamma(x, 10, 2), true = sim$values)

qdfauction documentation built on Nov. 10, 2026, 5:07 p.m.