qibm: Quantitative Imaging Biomarker Model

In brian-j-smith/qibm: Quantitative imaging biomarker modeling

Description

qibm is used to fit a Bayesian measurement error model for the comparison of biomarker measurements derived from different quantitative imaging methods.

Usage

qibm(fixed, image, operator, data, priors = list(), parameters = c("mu",
  "gamma.img", "Sigma.img", "sigma.opr", "sigma.imgopr", "sigma.err", "gof.rep",
  "gof.obs"), n.burnin = 2500, n.iter = 10000, n.thin = 5, n.chains = 3,
  seed = 123)

Arguments

fixed	fomula of the form y ~ method where y is a vector of biomarker measurements, or a transformation thereof, and method is a vector of identifiers for the imaging methods.
image	quoted or unquoted name of a vector of image identifiers.
operator	quoted or unquoted name of a vector of operator identifiers.
data	data.frame containing the data vectors.
priors	list of prior parameter values. See Details section below for list element names and defaults.
parameters	vector naming the model parameters to be returned (default: all).
n.burnin	number of MCMC samples to discard as a burn-in sequence.
n.iter	total number of samples to generate.
n.thin	period at which to save samples.
n.chains	number of parallel MCMC chains to generate.
seed	numeric random number generator seed.

Details

Prior distribution hyperprameters may be specified through the following list elements of the priors argument.

mu.mean

Normal mean for mu parameters (default: 0).

mu.var

Normal variance for mu paraemters (defaul: 1e6).

sigma2.opr.shape

Inverse-gamma shape for inter-operator variance parameters (default: 1e-3).

sigma2.opr.rate

Inverse-gamma rate for inter-operator variance parameters (default: 1e-3).

sigma.opr.lim

Two-element vector of uniform minimum and maximum for inter-operator standard deviation parameters (default: inverse-gamma distribution).

sigma2.imgopr.shape

Inverse-gamma shape for image-by-operator variance parameters (default: 1e-3).

sigma2.imgopr.rate

Inverse-gamma rate for image-by-operator variance parameters (default: 1e-3).

sigma.imgopr.lim

Two-element vector of uniform minimum and maximum for image-by-operator standard deviation parameters (default: inverse-gamma distribution).

sigma2.err.shape

Inverse-gamma shape for intra-opearator error variance parameters (default: 1e-3).

sigma2.err.rate

Inverse-gamma rate for intra-opearator error variance parameters (default: 1e-3).

sigma.err.lim

Two-element vector of uniform minimum and maximum for intra-opearator standard deviation parameters (default: inverse-gamma distribution).

Sigma.img.cor

Scalar pairwise correlation in the correlation matrix defining an inverse-Wishart scale matrix for between-image and method variances (default: 0).

Sigma.img.scale

Scalar multiple of the correlation matrix defining an inverse-Wishart scale matrix for between-image and method variances (default: 1).

Sigma.img.df

Inverse-Wishart degrees of freedom for between-image and method variances (default: number of methods).

Value

A qibm object that inherits from mcmc.list and contains the MCMC sampled model parameter values.

See Also

describe, with, Bias, CIndex, Cor, GOF, ICC, LRM RC, RDC, wCV.

Examples

## Not run: 
data(hnc)

## Default prior distributions
fit <- qibm(log(Volume) ~ method, image = lesion, operator = operator,
            data = hnc, n.burnin = 5000, n.iter = 10000, n.thin = 5,
            n.chains = 3)

describe(with(fit, exp(mu)))
describe(with(fit, exp(sqrt(diag(Sigma.img)))))
describe(with(fit, exp(sqrt(sigma.opr^2 + sigma.imgopr^2))))
describe(with(fit, exp(sigma.err)))

describe(Cor(fit))

describe(Bias(fit, log = TRUE))
describe(CIndex(fit))
describe(ICC(fit))

describe(wCV(fit, log = TRUE))
describe(RDC(fit))
describe(RC(fit))

fit.gof <- GOF(fit)
plot(fit.gof)
describe(fit.gof)


## User-specified Uniform(0, 2) priors on standard deviation parameters
fit2 <- qibm(log(Volume) ~ method, image = lesion, operator = operator,
             data = hnc, n.burnin = 5000, n.iter = 10000, n.thin = 5,
             n.chains = 3, priors = list(sigma.opr.lim = c(0, 2),
                                         sigma.imgopr.lim = c(0, 2),
                                         sigma.err.lim = c(0, 2)))

## End(Not run)

brian-j-smith/qibm documentation built on May 29, 2019, 2:12 p.m.