adaptiveDesign_normal: Simulate adaptive design where control sample size is...

In wiesenfa/ESS: Quantification of prior informativeness in terms of prior effective and effective current sample size

Description

Simulate adaptive design where control sample size is adapted according to prior effective sample size for normal outcome similar to Schmidli at al (2014).

Usage

adaptiveDesign_normal(ctl.prior, treat.prior, 
    N1, Ntarget, Nmin, M, 
    muc, mut, sc, st, sc.known=TRUE, st.known=TRUE,
    discard.prior = TRUE, 
    vague = mixnorm(c(1, 0, 10)), 
    ess = "ecss", ehss.method = "mix.moment", min.ecss, D=MSE,
    decision, 
    nsim = 100, cores = 1, seed = 123, progress="text")

Arguments

ctl.prior	RBesT normMix object (or powerprior object created by as.powerprior) as prior for the control group
treat.prior	RBesT normMix object (or powerprior object created by as.powerprior) as prior for the treatment group
N1	Sample size in each group at interim
Ntarget	Target sample size in control group
Nmin	Minimum number of samples in control group after interim analysis
M	Final sample size in treatment group
muc	True control mean
mut	True mean in treatment group
sc	standard deviation in control group
st	standard deviation in treatment group
sc.known	logical. If TRUE, assume sc to be known, otherwise replace by empirical standard deviation.
st.known	logical. If TRUE, assume st to be known, otherwise replace by empirical standard deviation.
discard.prior	Replace prior by vague prior if ESS<0?
vague	RBesT normMix object (single component mixture prior) serving as baseline vague prior
ess	either "ecss" or "ehss" for effective current or historical sample size, respectively.
ehss.method	if ess=="ehss". Specify version of EHSS as in ehss.
min.ecss	if ess=="ecss". Minimal ECSS of interest (negative). A large absolute value of min.ecss is computational expensive, could be set to -1 if discard.prior=TRUE and no interest in the ECSS estimate itself.
D	A function that measures informatives, e.g. MSE or user-specified function
decision	function created by decision2S.
nsim	Number of Monte Carlo iterations
cores	number of parallel cores used in mclapply
seed	random seed
progress	name of the progress bar to use (see create_progress_bar)

Details

The traditional approach to prior effective sample size (prior ESS) is aimed at quantifying prior informativeness, but is not aimed at detecting potential prior-data conflict.

The ECSS computes the prior effective sample size in terms of samples from the current data model (i.e., samples with characteristics consistent with the current trial). Under extreme prior-data conflict, the prior may account for a negative number of samples, showing that information is subtracted, rather than added, by the elicited prior. The ECSS quantifies the number of current samples to be added or subtracted to the likelihood in order to obtain a posterior inference equivalent to that of a baseline prior model (e.g. in terms of mean squared error, MSE). Fur further details, see Wiesenfarth and Calderazzo (2019).

Standard approach uses effective historical sample size (ess="ehss"), while Wiesenfarth and Calderazzo (2019) use the effective current sample size (ess="ecss"). When the ECSS is negative, the design provides the option of discarding the prior (discard.prior=TRUE).

Extensive documentation is given in the vignette.

Author(s)

Manuel Wiesenfarth

References

Schmidli, H., Gsteiger, S., Roychoudhury, S., O'Hagan, A., Spiegelhalter, D., and Neuenschwan- der, B. (2014). Robust meta-analytic-predictive priors in clinical trials with historical control information. Biometrics, 70(4):1023-103

Wiesenfarth, M., Calderazzo, S. (2019). Quantification of Prior Impact in Terms of Effective Current Sample Size. Submitted.

See Also

vignette("robustMAP",package="RBesT")

Examples

# see
# vignette("vignetteDesign", package = "ESS")

wiesenfa/ESS documentation built on June 19, 2019, 4:19 p.m.