Fit the Outcome Model

Sets up the model formulas and data for the pooled logistic regression and robust variance estimation and fits the models.

data_modelling(
  data,
  outcome_cov = ~1,
  model_var = NULL,
  first_followup = NA,
  last_followup = NA,
  use_weight = FALSE,
  analysis_weights = c("asis", "unweighted", "p99", "weight_limits"),
  weight_limits = c(0, Inf),
  use_censor = FALSE,
  include_followup_time = ~followup_time + I(followup_time^2),
  include_expansion_time = ~for_period + I(for_period^2),
  where_case = NA,
  glm_function = c("glm", "parglm"),
  use_sample_weights = TRUE,
  quiet = FALSE,
  ...
)

Arguments

data

A data.frame containing all the required columns.

outcome_cov

A RHS formula with baseline covariates to adjust in final model

model_var

List of Variables of interest to be used in final model. Derived variables to use in outcome models. Typically assigned_treatment for ITT and per-protocol, and dose + dose^2 for as-treated.

first_followup

First follow-up period

last_followup

Last follow-up period

use_weight

Use weights in analysis. If FALSE then no weights will be calculated.

analysis_weights

One of

• "asis": use the weights as calculated

• "p99": truncate weights at the 1st and 99th percentiles

• "weight_limits": truncate weights at the values specified in weight_limits

• "unweighted": set all analysis weights to 1, even with use_weight = TRUE

weight_limits

Lower and upper limits to truncate weights, given as c(lower, upper)

use_censor

Use censoring for per-protocol analysis - censor person-times once a person-trial stops taking the initial treatment value

include_followup_time

The model to include the follow up time of the trial (followup_time) in outcome model, specified as a RHS formula.

include_expansion_time

The model to include the trial period (for_period) in outcome model, specified as a RHS formula.

where_case

List of where conditions used in subsetting the data used in final analysis

glm_function

Which glm function to use for the final model from stats or parglm packages

use_sample_weights

Use sample weights in addition to IP weights. data must contain a column sample_weight. The weights used in the model are calculated as weight = weight * sample_weight.

quiet

Don't print progress messages.

...

Additional arguments passed to glm_function. This may be used to specify initial parameter estimates or arguments to control. See stats::glm, parglm::parglm and parglm::parglm.control() for more information.

Value

Object of class TE_model containing

• model, a glm object

• robust a list containing a coefficient summary table and the robust covariance matrix.

Details

The class variables parameters (outcomeClass,class_switchn, class_switchd,class_censen,class_censed) can be given as a character vector which will construct factors using as.factor or as a named list with the arguments for factor e.g. list(risk_cat=list(levels = c(1,2,3,0), age_cat=list(levels=c(1,2,3),labels=c("50-60","60-70","70+")