| Title: | Patient-Reported Outcome Data Analysis with Stan |
| Version: | 3.0.1.5 |
| Description: | Algorithms and subroutines for patient-reported outcome data analysis. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.1.2 |
| Biarch: | true |
| Depends: | R (≥ 3.5.0) |
| Imports: | methods, Rcpp (≥ 0.12.0), rstan (≥ 2.18.1), rstantools (≥ 2.1.1), BI |
| LinkingTo: | BH (≥ 1.66.0), Rcpp (≥ 0.12.0), RcppEigen (≥ 0.3.3.3.0), rstan (≥ 2.18.1), StanHeaders (≥ 2.18.0) |
| SystemRequirements: | GNU make |
| NeedsCompilation: | yes |
| Packaged: | 2025-11-07 18:22:21 UTC; alkb |
| Author: | Bin Wang  [aut,
cre] |
| Maintainer: | Bin Wang <bwang831@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2025-11-08 06:50:02 UTC |
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 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 |
|
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 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 |
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")
Changing point estimator to correct bias due to unblinding in RCTs
Description
estimate the sham effect and correct effect size
Usage
blinding.cpe(x,group,guess)
Arguments
x |
outcome variable. numeric vector |
group |
group assignment. Coded as "0"=control and "1"=active/treatment. If 'group' is a factor, the first level will be treated as "control" arm. For example, if there are two values (ie. "ctrl" and "active"), "active" will be treated as the control arm. If the two levels are "control" and "treatment", "control" will be treated as the control arm. |
guess |
responses to the blinding survey question. The response corresponding to positive sham effect needs to be coded as "1", and the rest as "0". If the possible responses are "Active", "Control" and "I don't know", code "guess" as "1" for "Active" and "0" for the others if a subject responded "active" is expect to have positive sham effect. Otherwise, if a subject responded "active" is expect to have negative sham effect, code "guess" as "0" for "Active" and "1" for the others. |
Details
To be added
Value
3 estimates, BI indices.
Examples
u1 = 5.5 # trt
u2 = 2.0 # ctrl
theta = 3.2 # sham
sigma2 = 2.5 # v(rij)
ntreat = 500
nsham = 500
beta0 = 1.0
beta1 = 2.0
beta2 = 1.0 # no contamination
Tind = c(rep(1, ntreat), rep(0,nsham)) #treatment group indicator
u1v = rep(u1,ntreat)
u2v = rep(u2,nsham)
uv = c(u1v,u2v)
tauv = uv - rep(u2, ntreat+nsham)
r = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma2))
q = 1/(1 + exp(-(beta0 + beta1*Tind + beta2*(tauv+r))))
bernGen = function(qq){rbinom(1,1,qq)}
I = sapply(q,bernGen)
x = uv + theta*I + r # fixed sham effect
## I have concerns about the error term(s). x.sham~N(theta,sigma.sham)?
sigma.sham = 1.5
r2 = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma.sham))
x = (uv + r) + theta*I #+ r2 # fixed sham effect
out1 <- blinding.cpe(x=x,group=Tind,guess=I);
out1
##data(bd012)
##blinding.cpe(x=bd012$y, group=bd012$group,guess=bd012$guess)
##data(bd011)
##blinding.cpe(x=bd011$y, group=bd011$group,guess=bd011$guess)
##data(bd010)
##blinding.cpe(x=bd010$y, group=bd010$group,guess=bd010$guess)
Latent Shift Logistic Regression
Description
To be updated.
Usage
blinding.test(x, group, guess, mu0 = 0, s0 = 1,...)
Arguments
x, guess |
outcome variable and guess response from blinding survey |
group |
group assignments. Current version support one or two groups only |
mu0, s0 |
initial mean and sd of the latent variable of having sham effects |
... |
Parameters ("adapt_delta","stepsize","max_treedepth") to improve model fitting/convergence. |
Value
'sig.sham': sd of effect size of sham treatment.
Examples
u1 = 5.5 # trt
u2 = 2.0 # ctrl
theta = 3.2 # sham
sigma2 = 2.5 # v(rij)
ntreat = 500
nsham = 500
beta0 = 1.0
beta1 = 2.0
beta2 = 1.0 # no contamination
Tind = c(rep(1, ntreat), rep(0,nsham)) #treatment group indicator
u1v = rep(u1,ntreat)
u2v = rep(u2,nsham)
uv = c(u1v,u2v)
tauv = uv - rep(u2, ntreat+nsham)
r = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma2))
q = 1/(1 + exp(-(beta0 + beta1*Tind + beta2*(tauv+r))))
bernGen = function(qq){rbinom(1,1,qq)}
I = sapply(q,bernGen)
x = uv + theta*I + r # fixed sham effect
## I have concerns about the error term(s). x.sham~N(theta,sigma.sham)?
sigma.sham = 1.5
r2 = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma.sham))
x = (uv + r) + theta*I #+ r2 # fixed sham effect
out1 <- blinding.test(x=x,group=Tind,guess=I);
out1
##data(bd012)
##blinding.test(x=bd012$y, group=bd012$group,guess=bd012$guess)
##data(bd011)
##blinding.test(x=bd011$y, group=bd011$group,guess=bd011$guess)
##data(bd010)
##blinding.test(x=bd010$y, group=bd010$group,guess=bd010$guess)
Simulated blinding data
Description
Simulated blinding survey datasets .
Format
Key variables
y | vector | outcome variable |
guess | vector | guess of treatments |
group | vector | group assignment |
Examples
data(bd012)
names(bd012)
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 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))