NEWS

---

riskRegression 2024.21.10

• [feature] add weight, vcov, confint and model.tables method to wglm for extracting IPCW, variance covariance matrix, coefficients with the confidence intervals and p-values.
• [internal] wglm automatically compute the variance covariance matrix (instead of doing it on the fly when calling summary). Can be disabled with argument se.
• [internal] speed-up for iidCox with non-unique times: it internals works on unique times and expand the results at the end to match the user-request.
• [internal] speed-up for predictCox with non-unique times when diag=TRUE: call iidCox with unique times to reduce memory usage.

riskRegression 2024.09.10

• add argument product.limit in predictCox and remove remove predictCoxPL. Side effect: while the influence function for the hazard is unchanged it the influence function relative to the survival scaled by the product limit survival (instead of exponential approximation).
• fix bug in how ties are handled in ate when using the double robust estimator.
• argument fitter in wglm can now take value "prodlim" to use the kaplan-meier or Aalen-Johensen estimator for the censoring weights.
• argument store.iid in predictCox and predict.CauseSpecificCox has been renamed store and can now be used to reduce memory usage.

riskRegression 2020.10.01

• modification of how ate stores the results: $riskComparison has been split in $diffRisk and $ratioRisk. Gformula is now refered as GFORMULA. Estimates for various estimators are now stores in different lines instead of columns.
• modification of the print function for ate to make it shorter.
• modification of the summary function for ate to have nicer looking tables.
• autoplot now support smoothing using monotone splines and can also display the first derivative of the spline.

riskRegression 2020.08.21

• calculation of the influence function is now more memory efficient in ate (argument store.iid="minimal")
• confint.ate now can display adjusted p-values (argument p.value=TRUE) and perform one- or two-sided tests (argument alternative).
• autoplot has been improved for predictCox, predict.CauseSpecificCox, and ate objects.

riskRegression 2018.06.13

• confint is now called by predictCox/predictCSC/ate by default. If the user wants to apply confint on the output of predictCox/predictCSC/ate, he should make sure that se or iid are exported.

riskRegression 2018.06.01

• NEW FEATURE: confint function computing confidence interval/bands, possibly using a transformation.

riskRegression 2018.04.05

• ate nows uses the boot package to compute p-values with bootstrap resampling

riskRegression 1.4.9

• predictCox can now compute baseline hazards for left and right censored data
• fix the influence function for the ate
• re-organize the test battery: check the baseline hazard in a separate file.

riskRegression 1.4.8

• re-organize the test battery
• fix the sign of the influence function for the survival

riskRegression 1.4.7

• add progress bar to ate when using foreach

riskRegression 1.4.6

• add the function InfluenceCoxTest comparing survival/risk obtained by two different models