sampling_gpu: Run a cmdstan model

In bstatcomp/project-cpath: R package for Temporal progression modelling using stan with GPU support

Description

Runs a cmdstan model. All parameters with the number 2 in the name are only used in case of 2 scores.

Usage

sampling_gpu(
  df,
  mod_name = "glm",
  SubjectIdVar = IDp,
  StudyIdVar = IDs,
  TimeVar = time,
  ScoreVar = score,
  is_pbo = placebo,
  CovariatesR = NULL,
  CovariatesB = NULL,
  m_r = 0,
  m_b = 0,
  ScoreVar2 = NULL,
  is_pbo2 = NULL,
  CovariatesR2 = NULL,
  CovariatesB2 = NULL,
  m_r2 = 0,
  m_b2 = 0,
  gpu_enabled = 1,
  init_list = NULL,
  num_samples = 1000,
  num_warmup = 1000,
  save_warmup = 0,
  thin = 1,
  engaged = 1,
  gamma = 0.05,
  delta = 0.8,
  kappa = 0.75,
  t0 = 10,
  init_buffer = 75,
  term_buffer = 50,
  window = 25,
  algorithm = "hmc",
  engine = "nuts",
  max_depth = 10,
  metric = "diag_e",
  metric_file = "",
  stepsize = 1,
  stepsize_jitter = 0,
  id = 0,
  seed = 1607674300,
  ...
)

Arguments

df	Input data frame.
mod_name	character. The name of the model. For now the only possible value is "glm".
SubjectIdVar	Name of the column which holds the subject IDs. Must be without quotation marks.
StudyIdVar	Name of the column which holds the study IDs. Must be without quotation marks.
TimeVar	Name of the column which holds the times. Must be without quotation marks.
ScoreVar	Name of the column which holds the scores (responses). Must be without quotation marks.
is_pbo	Name of the placebo column for score 1. Must be without quotation marks.
CovariatesR	formula. Formula of the form ~ A + B + ..., where the letters represent the names of columns of the covariates for rate.
CovariatesB	formula. Formula of the form ~ A + B + ..., where the letters represent the names of columns of the covariates for baseline.
m_r	binary. 0 if the covariates for rate are additive, 1 if they are multiplicative.
m_b	binary. 0 if the covariates for baseline are additive, 1 if they are multiplicative.
ScoreVar2	Name of the column which holds the second scores (responses) if applicable. Must be without quotation marks.
is_pbo2	Name of the placebo column for score 2. Must be without quotation marks.
CovariatesR2	formula. Formula of the form ~ A + B + ..., where the letters represent the names of columns of the covariates for rate.
CovariatesB2	formula. Formula of the form ~ A + B + ..., where the letters represent the names of columns of the covariates for baseline.
m_r2	binary. 0 if the covariates for rate are additive, 1 if they are multiplicative.
m_b2	binary. 0 if the covariates for baseline are additive, 1 if they are multiplicative.
gpu_enabled	binary. 1 to use the GPU model, 0 to use the CPU model.
init_list	list. The list of initialization parameters for the model.
num_samples	integer. The number of samples.
num_warmup	integer. The number of samples in the warmup phase of the sampling.
seed	integer. Random seed for the model.
...	Other parameters for the sampler.

Value

A list with 4 elements.

stan_model	An object of stanfit. Fitted model.
data_used	The exact data that stan model used.
maps	The maps from original subject and study IDs to the IDs used in the model.
pred_samples	The predicted values for each iteration.

bstatcomp/project-cpath documentation built on Feb. 26, 2020, 2:40 p.m.