poisson_lognormal: HMC sampling for Poisson log-normal mixed model

In ChristofSeiler/cytoeffect: Regression Models with Multivariate Outcomes for Mass Cytometry Experiments

Description

poisson_lognormal uses Hamiltonion Monte Carlo to sample form an extend Poisson log-normal mixed model. Each cell and protein marker has its own rate parameter following a linear model.

Usage

poisson_lognormal(
  df_samples_subset,
  protein_names,
  condition,
  group,
  r_donor,
  eta = 1,
  iter = 325,
  warmup = 200,
  num_chains = 1,
  adapt_delta = 0.8,
  seed = 1
)

Arguments

df_samples_subset	Data frame or tibble with proteins counts, cell condition, and group information
protein_names	A vector of column names of protein to use in the analysis
condition	The column name of the condition variable
group	The column name of the group variable
r_donor	Rank of the donor random effect covariance matrix
eta	Hyperparameter for LKJ prior
iter	Number of iteration per chain for the HMC sampler
warmup	Number of warm up steps per chain for the HMC sampler
num_chains	Number of HMC chains to run in parallel
adapt_delta	Parameter to control step size of numerical solver
seed	Set seed for HMC sampler (higher value means smaller step size, max is 1)

Value

A list of class cytoeffect_poisson containing

fit_mcmc	rstan object
protein_names	input protein names
condition	input condition variable
group	input group names
df_samples_subset	input df_samples_subset table

Examples

set.seed(1)
df = simulate_data(n_cells = 10)
str(df)
fit = poisson_lognormal(df,
                        protein_names = names(df)[3:ncol(df)],
                        condition = "condition",
                        group = "donor",
                        r_donor = 2,
                        warmup = 200, iter = 325, adapt_delta = 0.95,
                        num_chains = 1)

ChristofSeiler/cytoeffect documentation built on Nov. 24, 2020, 12:25 a.m.