Package 'chartreview'

Title: Adaptive Multi-Wave Sampling for Efficient Chart Validation
Description: Functionality to perform adaptive multi-wave sampling for efficient chart validation. Code allows one to define strata, adaptively sample using several types of confidence bounds for the quantity of interest (Lai's confidence bands, Bayesian credible intervals, normal confidence intervals), and sampling strategies (random sampling, stratified random sampling, Neyman's sampling, see Neyman (1934) <doi:10.2307/2342192> and Neyman (1938) <doi:10.1080/01621459.1938.10503378>).
Authors: Georg Hahn [aut, cre], Sebastian Schneeweiss [ctb], Shirley Wang [ctb]
Maintainer: Georg Hahn <[email protected]>
License: GPL (>= 2)
Version: 1.0
Built: 2025-01-17 14:44:19 UTC
Source: CRAN

Help Index

Bayesian credible interval for binomial quantity

Description

Bayesian credible interval for binomial quantity

Usage

credibleinterval(k, S, alpha)

Arguments

k

Number of experiments.
S

Observed number of successes.
alpha

Level.

Value

Bayesian credible interval.

References

.

Examples

require(chartreview)
print(credibleinterval(10,5,0.05))

Adaptive sampling algorithm which implements several types of sampling strategies

Description

Adaptive sampling algorithm which implements several types of sampling strategies

Usage

fullrun(
  dat1,
  S,
  dat2,
  mode = 1,
  batchsize = 100,
  raking = TRUE,
  rakingmode = 3,
  rakingthreshold = 0.05,
  sdEstimate = mad,
  minSamples = 10
)

Arguments

dat1

First dataset on which the strata are computed.
S

Matrix defining the strata.
dat2

Second dataset on which confidence intervals are computed.
mode

Sampling mode (1 for random sampling, 2 for stratified random sampling, 3 for Neyman's sampling).
batchsize

Batch size in each wave.
raking

Boolean flag to switch on raking.
rakingmode

Option for raking (1 for random sampling, 2 for deterministic allocation, 3 for residual resampling).
rakingthreshold

Threshold for applying raking to a stratum.
sdEstimate

The estimate of the standard deviation as a function handle (usually sd or mad).
minSamples

Minimum number of samples used in each iteration.

Value

List with the resampled datasets per wave.

References

.

Examples

require(chartreview)

Lai confidence sequence for binomial quantity

Description

Lai confidence sequence for binomial quantity

Usage

lai(n, x, alpha)

Arguments

n

Number of experiments
x

Observed number of successes.
alpha

Error probability.

Value

Binomial confidence interval.

References

Lai, TL (1976). On Confidence Sequences. Ann Statist 4(2):265-280.

Examples

require(chartreview)
print(lai(10,5,0.05))

Generate plots on confidence intervals and prediction

Description

Generate plots on confidence intervals and prediction

Usage

makeplot(
  dataset2,
  dat2,
  optionCI = 1,
  stopCI = NULL,
  alpha = 0.05,
  stoppingoption = 2,
  xlim = NULL,
  ylim = NULL,
  main = NULL,
  makePlot = TRUE
)

Arguments

dataset2

The output dataset of the function 'fullrun'.
dat2

Second dataset on which confidence intervals are computed, see function 'fullrun'.
optionCI

Parameter to switch between confidence intervals (1 for Lai's confidence bands, 2 for Bayesian credible intervals, 3 for normal confidence intervals).
stopCI

The stopping bounds.
alpha

The error used to compute confidence bands.
stoppingoption

Type of stopping criterion (1 for confidence interval included in stopCI, 2 for upper bound below or lower bound above stopCI, 3 for length restriction on confidence interval).
xlim

Optional parameter to set x-axis in plots.
ylim

Optional parameter to set y-axis in plots.
main

Optional parameter to set title of plots.
makePlot

Parameter to control plot output.

Value

List with confidence intervals (slot CIs), the stopping point (slot stopline), and the reason for stopping (stopreason, see function 'stoppingcriterion').

References

.

Examples

require(chartreview)

Normal confidence interval for continuous quantity

Description

Normal confidence interval for continuous quantity

Usage

normalci(x, a)

Arguments

x

Vector of samples.
a

Error probability.

Value

Normal confidence interval.

References

.

Examples

require(chartreview)
x <- rnorm(10)
print(normalci(x,0.05))

Different options for the stopping criterion

Description

Different options for the stopping criterion

Usage

stoppingcriterion(ci, stopCI, stoppingoption = 2)

Arguments

ci

Confidence interval as tuple vector.
stopCI

Either a confidence interval for stoppingoption=1 and stoppingoption=2, or a scalar for stoppingoption=3.
stoppingoption

Option to determine if the stopping criterion is satisfied (1 for confidence interval included in stopCI, 2 for upper bound below or lower bound above stopCI, 3 for length restriction on confidence interval).

Value

Boolean answer if stopping criterion reached.

References

.

Examples

require(chartreview)
stoppingcriterion(c(0.5,0.6), c(0.7,0.8), stoppingoption=1)

Statification of input data matrix into given strata

Description

Statification of input data matrix into given strata

Usage

stratum(x, S, index)

Arguments

x

Input data matrix.
S

Strata by row in matrix S, with 2 columns per variable aka startpoint [included] and endpoint [excluded].
index

Index of the stratum in S.

Value

Vector of indices belong to the given stratum

References

.

Examples

require(chartreview)
x <- matrix(runif(10),ncol=1)
strata <- (0:10)/10
S <- cbind(strata[-length(strata)],strata[-1])
print(stratum(x,S,1))

Check if some interval is a subset of another interval

Description

Check if some interval is a subset of another interval

Usage

subsetInterval(x, y)

Arguments

x

First interval given by tuple.
y

Second interval given by tuple.

Value

Boolean answer if "x subseteq y".

References

.

Examples

require(chartreview)
x <- sort(runif(2))
y <- sort(runif(2))
print(subsetInterval(x,y))