msm: Marginal structural model

In CICI: Causal Inference with Continuous (Multiple Time Point) Interventions

Marginal structural model

Description

Estimation of a marginal structural model using results from the parametric g-formula.

Usage

msm(X, formula, family = gaussian, se = NULL, cilevel = 0.95, abar=NULL)

Arguments

X	An object of class gformula produced by gformula, with option ret = TRUE.
formula	Form of the marginal structural model. Can be specified as a formula object, e.g., formula = VL.4 ~ efv.4, as a quoted expression, e.g., formula = quote(VL.4 ~ efv.4), or as a character string, e.g., formula = "VL.4 ~ efv.4".
family	A description of the error distribution and link function to be used in the model. See family for details of family functions.
se	A character string specifying the standard errors used to compute confidence intervals. One of c('bootstrap', 'glm'). See Details.
cilevel	Numeric value between 0 and 1 specifying the confidence level of the bootstrap confidence intervals. Defaults to 95%.
abar	Vector or matrix that is a subset of the intervention used in X. Can be used to fit an MSM on a subset of the stacked counterfactual data.

Details

The marginal structural model (MSM) is estimated as a GLM. Confidence intervals are calculated using GLM standard errors (if se = 'glm') or nonparametric boostrap standard errors (if se = 'bootstrap' and gformula was run with B > 0.) By default: se = 'bootstrap' if gformula was run with B > 0, and se = 'glm' otherwise.

Value

Returns a list of class msmResult:

MSM	The fitted MSM of class glm.
coefs	The estimated coefficients of the MSM.
CIlow	Lower confidence interval bounds for each coefficient.
CIup	Upper confidence interval bounds for each coefficient.
formula	The 'formula' input argument.
se	The 'se' input argument.
vcov	Covariance matrix.

See Also

gformula for estimating expected counterfactual outcomes under multiple intervention values.

Examples

data(EFV)
gf <- gformula(
  X = EFV, Anodes = c("efv.0", "efv.1", "efv.2", "efv.3", "efv.4"),
  Ynodes = c("VL.0", "VL.1", "VL.2", "VL.3", "VL.4"),
  Lnodes = c("adherence.1", "weight.1", "adherence.2", "weight.2",
             "adherence.3", "weight.3", "adherence.4", "weight.4"),
  abar = seq(0, 5), B = 10, ret = TRUE
)

msm(gf, VL.4 ~ efv.4, se = "bootstrap") # default if B>0
msm(gf, VL.4 ~ efv.4, se = "glm")       # fast, but not valid (undercoverage)

CICI documentation built on April 7, 2026, 5:08 p.m.