BMRMM {BMRMM}R Documentation

Bayesian Markov Renewal Mixed Models (BMRMMs)

Description

Provides inference results of both transition probabilities and duration times using BMRMMs.

Usage

BMRMM(
  data,
  num_cov,
  switch_off_random = FALSE,
  trans_cov_index = 1:num_cov,
  duration_type = "Continuous",
  duration_cov_index = 1:num_cov,
  duration_excl_prev_state = FALSE,
  duration_unit = NULL,
  duration_num_comp = 4,
  duration_init_shape = rep(1, duration_num_comp),
  duration_init_rate = rep(1, duration_num_comp),
  simsize = 10000,
  burnin = simsize/2
)

Arguments

data

a data frame containing – individual ID, covariate values, previous state, current state, duration times (if applicable), in that order

num_cov

total number of covariates provided in 'data'

switch_off_random

TRUE if only population-level effects are considered, default is FALSE

trans_cov_index

indices of covariates that are used for transition probabilities, default is all of the covariates

duration_type

one of 'None', 'Discrete', 'Continuous', default is 'Continuous'

duration_cov_index

indices of covariates that are used for duration times, default is all of the covariates plus the previous state

duration_excl_prev_state

TRUE if the previous state is excluded from duration times inference, default is FALSE

duration_unit

the discretization of duration times, only used when 'duration_type' is 'Discrete'

duration_num_comp

number of gamma mixture components for duration times, default is 4

duration_init_shape

initialization of mixture gamma shape parameters, default is a vector of 1 of size 'duration_num_comp'

duration_init_rate

initialization of mixture gamma rate parameters, default is a vector of 1 of size 'duration_num_comp'

simsize

total number of MCMC iterations, default is 10000

burnin

number of burn-ins for the MCMC iterations, default is simsize/2

Details

Returns list(trans_results, duration_results)

(1) Users have the option to ignore duration times or model duration times as a discrete or continuous variable via defining 'duration_type':
"None": ignores duration times
"Continuous": treat duration times as a continuous variable (DEFAULT)
"Discrete": treat duration times as a new state with discretization 'duration_unit'. If 'Discrete' is used, duration times becomes a new state type. 'duration_unit' must be specified if 'duration_type' is 'Discrete'. For example, if an duration time entry is 20 and 'duration_unit' is 5, then the model will add 4 consecutive new states. If an duration time entry is 23.33 and 'duration_unit' is 5, then the model will still add 4 consecutive new states as the blocks are calculated with the floor operation

(2) 'trans_results' contains the results of the transition probabilities:
- "Num_States": total number of states
- "Xexgns": covariates related to transition probabilities
- "dpreds": maximum level for each related covariate
- "MCMCparams": MCMC parameters: simsize, burnin and thinning factor
- "TP_Exgns_Post_Mean": posterior mean of transition probabilities for different combinations of exogenous predictors
- "TP_Exgns_Post_Std": posterior standard deviation of transition probabilities for different combinations of exogenous predictors
- "TP_Anmls_Post_Mean": posterior mean of transition distribution components for different individuals
- "TP_All_Post_Mean": posterior mean of transition distribution components for different combinations of exogenous predictors AND different individuals
- "TP_Anmls_Post_Std": standard deviation of transition probabilities among different mice
- "TP_Exgns_Diffs_Store": difference in posterior mean of transition probabilities for every pair of covariate levels given levels of the other covariates
- "TP_Exgns_All_Itns": exogenous transition probabilities for every MCMC iteration
- "Clusters": number of clusters for each covariate for each MCMC iteration
- "Type": an identifier for results, which is "Transition Probabilities".

(3) 'duration_results' contains the results of the duration times ('duration_results' is NULL if 'duration_type' is 'None'):
"K" <- number of gamma mixture components
"Xexgns" <- covariates related to duration times
"dpreds" <- maximum level for each related covariate
"MCMCparams" <- MCMC parameters: simsize, burnin and thinning factor
"Duration_Times" <- input duration times
"Comp_Assign" <- mixture component assignment for each data point in the last MCMC iteration
"Duration_Exgns_Store" <- posterior mean of mixture probabilities for different combinations of exogenous predictors of each MCMC iteration
"Marginal_Prob" <- estimated marginal mixture probabilities for each iteration
"Shape_Samples" <- estimated shape parameters for gamma mixtures for each iteration
"Rate_Samples" <- estimated rate parameters for gamma mixtures for each iteration
"Clusters" <- number of clusters for each covariate for each MCMC iteration
"Type" <- an identifier for results, which is "Inter-Syllable Intervals".

Value

List of results for transition probabilities and durations times, list(trans_results, duration_results). See details.

Author(s)

Yutong Wu, yutong.wu@utexas.edu

Examples


# In the examples, we use a shorted version of the foxp2 dataset, foxp2_sm

# ignores duration times and only models transition probabilities using all three covariates
results <- BMRMM(foxp2_sm,num_cov=2,duration_type='None',simsize=50)

# models duration times as a continuous variable with 5 gamma mixture components,
# using covariate 1 and the previous state
results <- BMRMM(foxp2_sm,num_cov=2,trans_cov_index=c(1),'duration_type'='Continuous',
                 duration_cov_index=c(1),duration_num_comp=5,simsize=50)

# models duration times as a discrete state with discretization 0.25 and
# do not include the previous state as a covariate
results <- BMRMM(foxp2_sm,num_cov=2,duration_type='Discrete',duration_excl_prev_state=TRUE,
                 duration_unit=0.025,simsize=50)



[Package BMRMM version 0.0.1 Index]