rbmi (version 1.6.1)

strategies: Strategies

Description

These functions are used to implement various reference based imputation strategies by combining a subjects own distribution with that of a reference distribution based upon which of their visits failed to meet the Missing-at-Random (MAR) assumption.

Usage

strategy_MAR(pars_group, pars_ref, index_mar)
strategy_JR(pars_group, pars_ref, index_mar)
strategy_CR(pars_group, pars_ref, index_mar)
strategy_CIR(pars_group, pars_ref, index_mar)
strategy_LMCF(pars_group, pars_ref, index_mar)

Value

A list with elements mu (a numeric vector of means) and sigma (a covariance matrix) giving the parameters of the subject's imputation distribution, formed by combining their own group's parameters (pars_group) with the reference group's parameters (pars_ref) according to the chosen strategy.

Arguments

pars_group

A list of parameters for the subject's group. See details.

pars_ref

A list of parameters for the subject's reference group. See details.

index_mar

A logical vector indicating which visits meet the MAR assumption for the subject. I.e. this identifies the observations after a non-MAR intercurrent event (ICE).

Details

pars_group and pars_ref both must be a list containing elements mu and sigma. mu must be a numeric vector and sigma must be a square matrix symmetric covariance matrix with dimensions equal to the length of mu and index_mar. e.g.

list(
    mu = c(1,2,3),
    sigma = matrix(c(4,3,2,3,5,4,2,4,6), nrow = 3, ncol = 3)
)

Users can define their own strategy functions and include them via the strategies argument to impute() using getStrategies(). That being said the following strategies are available "out the box":

• Missing at Random (MAR)

• Jump to Reference (JR)

• Copy Reference (CR)

• Copy Increments in Reference (CIR)

• Last Mean Carried Forward (LMCF)

Examples

# Parameters of the subject's own group and of the reference group
pars_group <- list(
    mu = c(1, 2, 3),
    sigma = as_vcov(c(1, 3, 2), c(0.4, 0.5, 0.45))
)
pars_ref <- list(
    mu = c(5, 6, 7),
    sigma = as_vcov(c(2, 1, 1), c(0.7, 0.8, 0.5))
)

# The first two visits meet the MAR assumption, the third does not
index_mar <- c(TRUE, TRUE, FALSE)

# Combine the parameters according to the different strategies
strategy_MAR(pars_group, pars_ref, index_mar)
strategy_JR(pars_group, pars_ref, index_mar)
strategy_CR(pars_group, pars_ref, index_mar)
strategy_CIR(pars_group, pars_ref, index_mar)
strategy_LMCF(pars_group, pars_ref, index_mar)