Specifications

The ‘‘Specifications’’ section gives detailed instructions about how to specify the required or optional arguments in different modules of pygformula to construct a specific analysis. To use the g-formula method in the package, the first step is to make sure that the input data meets the requirement of Input data. Then, users need to specify their parametric covariate model (see Covariate models), parametric outcome model (see Outcome model) , as well as the intervention of interest (see Interventions). Once these required modules are well-defined, the g-formula in pygformula can be called and output the results of the method.

Additionally, if there are censoring events, the package provides the option to obtain inverse probability weighted estimates for comparison with the g-formula estimates, see Censoring event. If there are competing events, the package provides two options for handling competing events in the case of survival outcomes, see Competing event. The package also provides option for calculating the hazard ratio of any two interventions of interest in Hazard ratio. If the data structure contains visit process, users can also perform g-formula analysis for this setting in Visit process. If there is deterministic knowledge about the relationship between the variables, it can be incorporated into the estimation of g-formula by applying restrictions, see Deterministic knowledge.

Contents:

• Input data
• Interventions
  • Natural course intervention
  • Static interventions
  • Dynamic interventions
  • Threshold interventions
• Covariate models
  • Functions of covariate histories
  • Different covariate distributions
• Outcome model
  • Survival outcome
  • Binary end of follow-up outcome
  • Continuous end of follow-up outcome
  • Custom model
• Censoring event
• Competing event
• Hazard ratio
• Visit process
• Deterministic knowledge
  • Restrictions on covariates
  • Restrictions on outcome
  • Restrictions on competing event
• Output
  • Numerical results
  • Graphical results