testSmallScales {stepR}

Test Small Scales

Description

For developers only; users should look at the function improveSmallScales in the CRAN package clampSeg. Implements the second step of HILDE (Pein et al., 2020, Section III-B) in which an initial fit is tested for missed short events.

Usage

.testSmallScales(data, family, lengths = NULL, q, alpha, ...)

Arguments

data

a numeric vector containing the observations

family

a string specifying the assumed parametric family, currently "LR" and "2Param" are supported

lengths

an integer vector giving the set of lengths, i.e. only intervals of these lengths will be considered. By default (NULL) 1:20 will be used for parametric family "LR" and 1:65 will be used for parametric family "2Param"

q

either NULL, then the vector of critical values at level alpha will be computed from a Monte-Carlo simulation or a numeric vector giving the vector of critical values. Either q or alpha must be given. Otherwise, alpha == 0.5 is chosen with a warning. This argument will be passed to critVal to obtain the needed critical values. Additional parameters for the computation of q can be specified in ..., for more details see the documentation of critVal. Please note that by default the Monte-Carlo simulation will be saved in the workspace and on the file system, for more details see Section Storing of Monte-Carlo simulations below

alpha

a probability, i.e. a single numeric between 0 and 1, giving the significance level. Its choice is a trade-off between data fit and parsimony of the estimator. In other words, this argument balances the risks of missing change-points and detecting additional artefacts. For more details on this choice see (Frick et al., 2014, section 4) and (Pein et al., 2017, section 3.4). Either q or alpha must be given. Otherwise, alpha == 0.5 is chosen with a warning

...

there are two groups of further arguments:

1. further parameters of the parametric family,

2. further parameters that will be passed to critVal. critVal will be called automatically with the number of observations n = length(y), the arguments family, intervalSystem, lengths, q and output set. For these arguments no user interaction is required and possible, all other arguments of critVal can be passed additionally

Value

a list with entries jumps, addLeft, addRight, noDeconvolution, data, q

References

Pein, F., Bartsch, A., Steinem, C., and Munk, A. (2020) Heterogeneous idealization of ion channel recordings - Open channel noise. Submitted.