v0.7.3

• Added Subgroup Treatment Effect Pattern (STEP) model fitting functions fit_rsp_step and fit_survival_step, the corresponding tidy method tidy.step as well as the graph function g_step.
• Added new layout function compare_vars which compares variables of different types between columns and produces a p-value for the comparison to the reference column. It is built on top of the summarize_vars functionality.
• Added utility functions:
  • cut_quantile_bins cuts a numeric vector into quantile bins.
  • fct_collapse_only collapses levels of a factor and only keeps those in the order provided.
  • fct_explicit_na_if inserts explicit missings in a factor based on a condition.
  • range_noinf is a kind of a wrapper function of base::range. It returns c(NA, NA) instead of c(-Inf, Inf) for zero-length data.
• Cox regression via fit_coxreg_univar and fit_coxreg_multivar is now also possible without treatment arm. In the univariate case this means that separate univariate models for the provided covariates are fitted and the corresponding effect estimates can later be tabulated.
• Added fraction in result returned by s_count_occurrences. It contains a list of numerators and denominators with one element per occurrence.
• Updated sum_num_patients and count_occurrences for the result unique and count_fraction to return (0, 0) when input is empty.
• Added a new argument groups_lists to extract_survival_subgroups, extract_rsp_subgroups and associated helper functions which allows to group factor levels of subgroup variables into manually defined groups, enhancing the flexibility of the resulting forest graphs.
• Forest graph function g_forest now extracts default arguments from attributes of the input table produced by tabulate_rsp_subgroups and tabulate_survival_subgroups so that the user does not have to do this manually anymore.
• In g_km:
  • Remove arm variable name from arm labels in plot legend and annotation tables.
  • Show symbol used to mark censored cases and match order of arm variable factor levels in the legend.
  • Display hazard ratio and its confidence interval to two decimal places.
  • Updated default position of hazard ratio table to stay on the left bottom corner but above x-axis.
  • Use arm value as strata name in when treatment arm variable only has one level.
• Updated s_surv_time function to use a newly created function range_noinf instead of base::range.
• New argument no_fillin_visits added to h_adsl_adlb_merge_using_worst_flag to specify the visits that are excluded from post-baseline worst toxicity grade output. Improved h_adsl_adlb_merge_using_worst_flag to include variables shared between adsl and adlb, along with PARAM, PARAMCD, ATOXGR, BTOXGR and optionally AVISIT, AVISITN when by_visit = TRUE. Previously, output only contains USUBJID, ARMCD, PARAMCD, ATOXGR, and BTOXGR.
• Fix bug in s_surv_timepoint for cases when there are zero patients at risk.
• Modified stat_median_ci function so that when empty var with empty name is passed, no row names contain missing values error would show.
• Deprecated s_cox_univariate function, use fit_coxreg_univar function instead.
• Updated default formats of hr and hr_ci in a_coxph_pairwise and median in s_surv_time to align with STREAM.
• Updated the pre-processing code in the files test-table_ttet01.R and test-table_dort01.R to make sure the analysis variable EVNT1 has both levels of the factor defined.
• Improved error message when number of levels in a factor variable in a data frame is not as expected.