Our approach
In this paper, we discuss a method of “repeated bootstrapping” (rBS)
where a number of iterations is chosen, and the resampling is repeated a
specified number of times. In the traditional bootstrapping method, a
diagnosis is made after a specified number of iterations (e.g. 100
iterations). However, if the test were to be repeated, the same
diagnosis may not be recommended given the random sampling involved in
the bootstrapping procedure11Note that this
limitation is also relevant for any number of diagnostic tests that
are performed on the same individual at different points in time or by
different services or technologies.. When considering the efficacy
of a diagnostic test it is vital to report its precision, or how similar
the results of the test are upon re-testing. To our knowledge, no
previous study has evaluated the repeatability of the 100 iterations
bootstrap test in the P300-based concealed information detection
literature, so that is what we aim to do here. Additionally, the rBS
technique could be a useful tool to calculate confidence intervals for
diagnostic metrics like the BSITER score, improving upon standard
methods in our field similar to ERP data quality metrics such as the one
recently proposed by Luck et al. (2021).