Package 'prome'

The 'prome' package.

Description

Algorithms to implenment the Bayesian methods to denoise the measurement errors in patient-reported outcome data with repeated measures. Also, two algorithms are included to discount the subgroup means or proportions for clinical studies with multiple subgroups.

---

Bayesian Hierarchical Model for RPO data with repeated measures

Description

A Bayesian hierachical model to denoise PRO data using repeated measures.

Usage

bate(x0,x1,group,z,x.range,...)
ResponderAnalysis(x,mcid,type="absolute",conf.level=0.95,show=TRUE)

Arguments

x0, x1	Numeric vector/matrix of observations at T0 (baseline) and T1 (end point) of a study.
z	covariates
group	group assignments. Current version support one or two groups only
x.range	range of data 'x0' and 'x1'
x	An R object generated by memixed
mcid	A threshold to define 'responder'
type	The type of responder analysis: absolute or relative changes
conf.level	Confidence level of the credible interval
show	control whether results should be displayed
...	Parameters ("adapt_delta","stepsize","max_treedepth") to improve model fitting/convergence.

Value

• 'xfit': fitted results using stan.

• 'mu.t0': baseline mean.

• 'sig.t0': baseline SD.

• 'sig.me': SD of measurement errors.

• 'mu.active': mean effect size of active treatment.

• 'sig.active': sd of effect size of active treatment.

• 'mu.sham': mean effect size of sham treatment.

• 'sig.sham': sd of effect size of sham treatment.

Examples

data(n100x3)
out1  <-  bate(x0=ex100x3$w0,x1=ex100x3$w1,group=ex100x3$group)
out1
ResponderAnalysis(out1,mcid=1,type="abs")
out2  <-  bate(x0=ex100x3$w0,x1=ex100x3$w1,group=ex100x3$group,
    control = list(adapt_delta = 0.8,
               stepsize = 5,
               max_treedepth = 10)
)
out2
ResponderAnalysis(out2,mcid=1,type="abs")
out <- out2
ResponderAnalysis(out,mcid=0.5,type="abs")
ResponderAnalysis(out,mcid=1,type="abs")
ResponderAnalysis(out,mcid=1.5,type="abs")
ResponderAnalysis(out,mcid=0.3,type="relative")
ResponderAnalysis(out,mcid=0.2,type="relative")
ResponderAnalysis(out,mcid=0.1,type="relative")

---

Sample PRO Data With Repeated Measures

Description

A simulated data set of patient-reported outcomes with repeated measures.

Format

A data frame with observations at beaseline and at a follow-up time.

w0	matrix	measures at baseline
w1	matrix	measures at follow-up time
group	character	group assignment

---

Bayesian Hierarchical Model for Information Borrowing for Means

Description

To compute the mean values of subgroups based on a Bayesian hierarchical model.

Usage

MeanHM(x,sigma)

Arguments

x	Numeric vector of observations for the subgroups.
sigma	hyper-parameter. to be estimated or can be given.

Value

• 'theta': population mean.

• 'sigma': population standard deviation.

Examples

x1 <- rnorm(100,2,1)
x2 <- rnorm(100,3,1.5)
x3 <- rnorm(100,4,1.9)
x <- cbind(x1,x2,x3)
MeanHM(x,sigma=0.5)

---

Bayesian Hierarchical Model for Information Borrowing for Proportions

Description

To compute the proportions of the subgroups assuming the subgroups follow the same binomial distribution with parameter p. The approach on partial pooling by Bob Carpenter has been used – "Hierarchical Partial Pooling for Repeated Binary Trials" https://mc-stan.org/users/documentation/case-studies/pool-binary-trials.html

Usage

PropHM(x, n, kappa)

Arguments

x	Numeric vector of events.
n	Numberic vector of group sample sizes.
kappa	kappa=alpha+beta>1. Must be given if the number of subgroups is 2.

Value

• 'data': data with estimates.

• 'alpha': parameter of the beta distribution.

• 'beta': parameter of the beta distribution.

Examples

out <- PropHM(x=c(5,10,2),n=c(20,50,30))

---

Bayesian analysis of 2x2 crossover trial data

Description

A Bayesian hierachical model to analysis data from 2x2 (AB/BA) crossover trials.

Usage

xover(group,y1,y2,y0,...)

Arguments

y0, y1, y2	vectors of data from baseline, period 1, and period 2, respectively.
group	group or treatment sequence.
...	other parameters, i.e. 'control' for model fitting.

Value

• 'stat': summary statistics.

• 'best': estimates using Bayesian analysis.

Examples

xover(y0=rnorm(20,34,1.5),y1=rnorm(20,30,2),
         y2=rnorm(20,25,1.5),group=round(runif(20)<0.5))

---