robustify: Robustify Mixture Priors

Description Usage Arguments Details Value Methods (by class) References See Also Examples

View source: R/robustify.R

Description

Add a non-informative component to a mixture prior.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
robustify(priormix, weight, mean, n = 1, ...)

## S3 method for class 'betaMix'
robustify(priormix, weight, mean, n = 1, ...)

## S3 method for class 'gammaMix'
robustify(priormix, weight, mean, n = 1, ...)

## S3 method for class 'normMix'
robustify(priormix, weight, mean, n = 1, ..., sigma)

Arguments

priormix

orior (mixture of conjugate distributions).

weight

weight given to the non-informative component (0 < weight < 1).

mean

mean of the non-informative component. It is recommended to set this parameter explicitly.

n

number of observations the non-informative prior corresponds to, defaults to 1.

...

optional arguments are ignored.

sigma

Sampling standard deviation for the case of Normal mixtures.

Details

It is recommended to robustify informative priors derived with gMAP using unit-information priors . This protects against prior-data conflict, see for example Schmidli et al., 2015.

The procedure can be used with beta, gamma and normal mixture priors. A unit-information prior (see Kass and Wasserman, 1995) corresponds to a prior which represents the observation of n=1 at the null hypothesis. As the null is problem dependent we strongly recommend to make use of the mean argument accordingly. See below for the definition of the default mean.

The weights of the mixture priors are rescaled to (1-weight) while the non-informative prior is assigned the weight given.

Value

New mixture with an extra non-informative component named robust.

Methods (by class)

References

Schmidli H, Gsteiger S, Roychoudhury S, O'Hagan A, Spiegelhalter D, Neuenschwander B. Robust meta-analytic-predictive priors in clinical trials with historical control information. Biometrics 2014;70(4):1023-1032.

Kass RE, Wasserman L A Reference Bayesian Test for Nested Hypotheses and its Relationship to the Schwarz Criterion J Amer Statist Assoc 1995; 90(431):928-934.

See Also

mixcombine

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
bmix <- mixbeta(inf1=c(0.2, 8, 3), inf2=c(0.8, 10, 2))
plot(bmix)
rbmix <- robustify(bmix, weight=0.1, mean=0.5)
rbmix
plot(rbmix)

gmnMix <- mixgamma(inf1=c(0.2, 2, 3), inf2=c(0.8, 2, 5), param="mn")
plot(gmnMix)
rgmnMix <- robustify(gmnMix, weight=0.1, mean=2)
rgmnMix
plot(rgmnMix)

nm <- mixnorm(inf1=c(0.2, 0.5, 0.7), inf2=c(0.8, 2, 1), sigma=2)
plot(nm)
rnMix <- robustify(nm, weight=0.1, mean=0, sigma=2)
rnMix
plot(rnMix)

RBesT documentation built on Nov. 24, 2021, 5:07 p.m.