Skip to contents

Fit outcome model

Source: R/fitModel.R
fitModel.Rd

Fits weighted marginal outcome model as a generalized linear model of the user's choosing, relating exposure main effects to outcome using IPTW weights.

Usage

fitModel(
  data,
  obj,
  weights = NULL,
  outcome,
  model = "m0",
  int_order = NA,
  covariates = NULL,
  family = gaussian(),
  link = "identity",
  verbose = FALSE,
  save.out = FALSE
)

# S3 method for class 'devMSM_models'
print(x, i = NA, save.out = FALSE, ...)

Arguments

data

data in wide format as: a data frame, list of imputed data frames, or mids object from the mice package

obj

initialized MSM object from initMSM()

weights

list of IPTW weights output from createWeights()

outcome

name of outcome variable with ".timepoint" suffix. See initMSM() for details on suffix

model

character indicating one of the following outcome models:

  • "m0" (exposure main effects)

  • "m1" (exposure main effects & covariates)

  • "m2" (exposure main effects & their interactions)

  • "m3" (exposure main effects, their interactions, & covariates)

int_order

integer specification of highest order exposure main effects interaction, required for interaction models ("m2", "m3")

covariates

list of characters reflecting variable names of covariates, required for covariate models ("m1", "m3")

family

(optional) family function specification for WeightIt::glm_weightit() model. Note that this should be specified as as a function, not a character string unless you have a multinomial outcome, in which case set this to "multinomial", or if you have an ordinal outcome with more than 2 levels, in which case set this to "ordinal". (default is gaussian)

(optional) link function specification for WeightIt::glm_weightit() model.

verbose

(optional) TRUE or FALSE indicator for printing output to console. default is FALSE.

save.out

(optional) Either logical or a character string. If TRUE, it will output the result to a default file name within home_dir set in initMSM(). You can load the data with x <- readRDS(file). To use a non-default file name, specify a character string with the file name. It will save relative to home_dir. There might be naming conflicts where two objects get saved to the same file. In these cases, users should specify a custom name. default is FALSE.

x

devMSM_models object from fitModel

i

For multiply imputed datasets, i selects which imputation to print results for. Default is i = 1. With i = TRUE, all imputed datasets will be looped over. With i = NULL, will average over all imputed datasets and summarize that. Ignored for non-imputed data.

...

ignored

Value

list containing WeightIt::glm_weightit() model output. It is the length of the number of datasets (1 for a data.frame or the number of imputed datasets)

See also

WeightIt::glm_weightit() for more on family/link specifications.

Examples

library(devMSMs)
data <- data.frame(
  ID = 1:50,
  A.1 = rnorm(n = 50),
  A.2 = rnorm(n = 50),
  A.3 = rnorm(n = 50),
  B.1 = rnorm(n = 50),
  B.2 = rnorm(n = 50),
  B.3 = rnorm(n = 50),
  C = rnorm(n = 50),
  D.3 = rnorm(n = 50)
)
obj <- initMSM(
  data,
  exposure = c("A.1", "A.2", "A.3"),
  ti_conf = c("C"),
  tv_conf = c("B.1", "B.2", "B.3", "D.3")
)
f <- createFormulas(obj, type = "short")
w <- createWeights(data = data, formulas = f)

fit_m0 <- fitModel(
  data = data, weights = w, 
  outcome = "D.3", model = "m0"
)
#> Error in all_fits[[1]]: subscript out of bounds
print(fit_m0)
#> Error: object 'fit_m0' not found

fit_m1 <- fitModel(
  data = data, weights = w, 
  outcome = "D.3", model = "m1", 
  covariates = c("C")
)
#> Error in all_fits[[1]]: subscript out of bounds
print(fit_m1)
#> Error: object 'fit_m1' not found

fit_m2 <- fitModel(
  data = data, weights = w, 
  outcome = "D.3", model = "m2", 
  int_order = 2
)
#> Error in all_fits[[1]]: subscript out of bounds
print(fit_m2)
#> Error: object 'fit_m2' not found

fit_m3 <- fitModel(
  data = data, weights = w, 
  outcome = "D.3", model = "m3",
  int_order = 2, covariates = c("C")
)
#> Error in all_fits[[1]]: subscript out of bounds
print(fit_m3)
#> Error: object 'fit_m3' not found

data <- data.frame(
  ID = 1:50,
  A.1 = rnorm(n = 50),
  A.2 = rnorm(n = 50),
  A.3 = rnorm(n = 50),
  B.1 = rnorm(n = 50),
  B.2 = rnorm(n = 50),
  B.3 = rnorm(n = 50),
  C = rnorm(n = 50),
  D.3 = c(rep(c("A", "B", "C"), 16), "A", "B")
)
obj <- initMSM(
  data,
  exposure = c("A.1", "A.2", "A.3"),
  ti_conf = c("C"),
  tv_conf = c("B.1", "B.2", "B.3", "D.3")
)
f <- createFormulas(obj, type = "short")
w <- createWeights(data = data, formulas = f)

fit_m0 <- fitModel(
  data = data, weights = w, 
  outcome = "D.3", model = "m0", family = "multinomial"
)
print(fit_m0)
#> Error: There are duplicate labels in the estimates table. Call `get_estimates()` to see the available identifiers, and use the `shape` argument to specify the (potentially nested or grouped)  nature of parameters. Alternatively, duplication can happen when you use `coef_map` or `coef_rename` to assign the same label to multiple variables, but these variables are part of the same model. Please check your `coef_map` and `coef_rename` arguments.