bca_coverage: Estimate the number of BCa bootstrap replicates
In elipickh/rptGam: Repeatability estimates for random effect GAM models

Description Usage Arguments Details Value Author(s) References

View source: R/bca_coverage.R

Estimate the number of BCa bootstrap replicates

bca_coverage(
  A,
  ci = 0.95,
  perc_dev = 10,
  prob = 0.95,
  boots = NULL,
  point = NULL
)

`A`	A numeric value, indicating the acceleration factor. Can be estimated from the data (i.e., using jackknife) or, if A is bnull, will be estimated from boots as in coxed::bca, but using the point estimate instead of mean of boots in the calcualtion.
`perc_dev`	A numeric in (0, 100), indicating the maximum percentage deviation of each of the upper and lower confidence interval lengths. The lower this value, the more boostraps are typically required.
`point`	A numeric value. The sample statistic from the original dataset.
`prob:`	A numeric n (0,1), indicating the approximate probability that perc_dev holds.

"The accuracy obtained by a given choice of B is random, because the bootstrap simulations are random. To determine an appropriate value of B, we specify a bound on the percentage deviation, denoted pdb, and we require that the actual percentage deviation is less than this bound with a specified probability, 1 - r, close to one. The three-step method takes pdb and r as given and specifies a data-dependent method of determining a value of B...such that the desired level of accuracy is achieved. For example, one might take (pdb, r) = (10, .05). In this case, the three-step method determines a value B* such that the percentage deviation of the upper and lower confidence interval lengths is less than 10 approximate probability .95." (Andrews & Buchinsky, 2002) Authors recommend iterating through this process for (possibly) better finite accuracy.

An integer, indicating the estimated number of bootstrap replicates. This value is either the initial estimate (if boots is null) or the final estimate (if boots is not null).

Eliezer Pickholtz (eyp3@cornell.edu)

Andrews, D. W., & Buchinsky, M. (2002). On the number of bootstrap repetitions for BCa confidence intervals. Econometric Theory, 18(4), 962-984.

elipickh/rptGam documentation built on Nov. 25, 2019, 10:08 a.m.