- Fixed an error for joint interventions on multiple treatments
- Fixed an error occurring when multiple restrictions are applied to a single variable
- Revised the
`gformula()`

function so that it produces a warning message rather than an error message when one of the bootstrap replicates fails. The bootstrap standard errors and 95% CIs are calculated based on the bootstrap replicates that do not fail. - Fixed an error occurring when no interventions are supplied (i.e., only the natural course intervention is used)
- Slightly sped up the calculation of the counterfactual cumulative risks
- Expanded the error checking

- Fixed an error in the
`gformula()`

function that assumed that the name of the ID variable in`obs_data`

was`'id'`

- Removed Travis CI

- Revised the plot of the estimates of the natural course risk so that it starts at (0, 0)
- Fixed an error when obtaining confidence intervals around the hazard ratio estimates
- Fixed an error in the reported standard errors of the coefficients of the fitted categorical covariate models
- Fixed an error in the reported root mean squared error values for the outcome and competing event models
- Allowed categorical covariates to be of class “numeric” (rather than requiring them to be of class “factor”)

- Added the “cumulative percent intervened on” and “average percent intervened on” to the output of the
`gformula()`

function - Added option for users to carry forward the natural value of treatment rather than the intervened value. See the
`int_visit_type`

argument in the`gformula()`

function - Added option for users to access the bootstrap replicates of the parametric g-formula estimates. See the
`boot_diag`

argument in the`gformula()`

function. - Fixed an error in computing the inverse probability weighted means of the time-varying covariates

- Added option for users to specify censoring models to compute inverse probability weights for estimating the natural course means / risk from the observed data
- Added data set
`censor_data`

and a corresponding example application in the documentation to illustrate the application of inverse probability weighting for estimating the natural course means / risk from the observed data - Fixed an error in calculating the means of the time-varying covariates under the natural course for survival outcomes
- Fixed errors in calculating the observed risk estimates and g-formula survival estimates when competing events are not treated like censoring events
- For categorical time-varying covariates, the
`plot.gformula_survival()`

,`gformula_continuous_eof()`

, and`gformula_binary_eof()`

functions now display the nonparametric/IP weighted and parametric g-formula estimates of the probability of observing each level of the covariate. Previously, these functions displayed the counts of categorical variables.

- Updated computation of (lagged) cumulative averages to use the recursive formula. There should be a noticeable improvement in the computation time when using several (lagged) cumulative average terms and when the number of time points is large.
- Fixed an error for covariates of type
`truncated normal`

(Thanks to - Updates to the documentation

- Fixed error in the
`coef.gformula()`

example

- Added wrapper function called
`gformula()`

for the`gformula_survival()`

,`gformula_continuous_eof()`

, and`gformula_binary_eof()`

functions. Users should now use the more general`gformula()`

function to apply the g-formula. - Added option for users to specify the values for lags at pre-baseline times by including rows at time -1, -2, …, -i.
- Added an example data set called
`continuous_eofdata_pb`

, which illustrates how to prepare a data set with pre-baseline times - Added option for users to pass in “control parameters” (e.g., maximum number of iterations, maxit, in glm.control) when fitting models for time-varying covariates via the
`covparams$control`

argument. (Thanks to @jerzEG for the suggestion) - Added option for users to access the fitted models for the time-varying covariates, outcome, and competing event (if applicable). See
`model_fits`

argument of the`gformula()`

function - Added simulated data under the natural course to the
`sim_data`

component of the output of the`gformula()`

function - Added a progress bar for the number of bootstrap samples completed. See the
`show_progress`

argument of the`gformula()`

function for further details - Added
`summary()`

,`coef()`

, and`vcov()`

S3 methods for objects of class ‘gformula’ - Added argument
`fits`

in the`print.gformula_survival()`

,`print.gformula_continuous_eof()`

, and`print.gformula_binary_eof()`

functions. Added argument`all_times`

in the`print.gformula_survival()`

function - Fixed minor bug in the
`lagavg()`

function - Fixed bug occuring when not using lags of the intervention variable(s)
- Fixed bug occuring in the truncation beyond covariate ranges. (Thanks to Louisa Smith)
- Updates to the documentation

- First version released on CRAN (https://CRAN.R-project.org/package=gfoRmula)
- Updates to the documentation

- Removed
`example_intervention1()`

,`example_intervention2()`

, and`visit_sum_orig()`

, as these functions are not used internally and users should not directly apply them - Removed export of
`visit_sum()`

and`natural()`

, as these functions are used internally and users should not directly apply them - Updates to the documentation

- Minor updates to the documentation

- First version released on GitHub (https://github.com/CausalInference/gfoRmula)