innerBLB: Bag of little bootstraps

In schwenzfeier/udp: Statistically Useful Differential Privacy

Description

Implements the bag of little bootstraps algorithm (Kleiner et al., 2014). Splits the dataset and calculates the weighted QOI on each partition, also calculates alpha_1 and alpha_2

Usage

innerBLB(x, statistic, metric, B, N, lambda, ...)

Arguments

x	Dataset partition
statistic	Function that calculates quantity of interest
metric	Summary metric (default is mean and variance)
B	Number of bootstrap simulations
N	Dataset n-size
lambda	Bounding parameter for the QOI
...	Parameters required for statistic

Value

Returns a list:

metrics	Summary metric across bootstraps: default is mean and variance
a_2	Proportion of bootstrapped QOIs that are above lambda
a_1	Proportion of bootstrapped QOIs that are above -lambda
res	Vector of B bootstrapped QOIs

schwenzfeier/udp documentation built on April 1, 2021, 3:39 a.m.