NEWS.md
In drugDemand: Drug Demand Forecasting

drugDemand 0.1.3

remove the dependence on pscl and icenReg packages
add stringr to imported packages
add survival::survreg and stats::optim to imported functions
use the Brent method for optimization to fit an interval-censored exponential distribution to the time from randomization to the first drug dispensing visit
check the number of unique combinations of time and k1 in f_fit_ti
use the VarCorr function from the nlme package to obtain the variance of random effects from the lme function call
use data from all kit types to inform the common time model
add the pred_pp_only parameter to f_drug_demand to make protocol-based predictions only
ensure that df and visitview have all the required columns and none of the required columns have missing values
combine dosing_pred_df and dosing_pred_pp in the f_drug_demand output
handle cases where all patients in a treatment arm discontinued before the cutoff
replace round with formatC to retain the zeros after the decimal point
rename fit_xx to xx_fit, where xx = k0, t0, t1, ki, ti, di to be consistent with enroll_fit and event_fit naming convention
add kit and kit_name to handle cases when a drug has different kit types depending on the site
add the prior probability of different kit types within a drug at the design stage
streamline the examples through the use of function calls
update the initial parameter values in f_fit_ki for zero-inflated Poisson model
update the p.fit calculation in f_fit_ti when using the least absolute deviations model
convert kit_name to a factor to ensure the correct order when creating a plot
add kit_name as the sub plot title in f_dispensing_models.R
add the vf_kit parameter to f_dose_draw_1 to get the kit information for each subject in each simulation draw
add the f_ongoing_new function to prepare the dosing data sets to impute for ongoing and new subjects
add kit_description_df to the output of f_dose_observed
add the rdirichlet function to generate cell probabilities from the Dirichlet distribution
remove the f_treatment_by_drug_df function
replace the treatment_by_drug matrix with the treatment_by_drug_df data frame
add kit_description_df, treatment_by_drug_df, and dosing_schedule_df to the f_drug_demand output

drugDemand 0.1.2

add least absolute deviations (LAD) regression as an option for modelling the gap time between drug dispensing visits and rename the original linear model as least squares (LS)
use more descriptive names for drug dispensing models

drugDemand 0.1.1

add a reference for parametric analysis of interval-censored survival data
only keep one record per subject and drug dispensing day when using a common time model
rename f_dosing_draw, f_dosing_draw_1, and f_dosing_draw_t_1 to f_dose_draw, f_dose_draw_1, and f_dose_draw_t_1, respectively
remove the vf_new parameter of the f_dose_draw_1 function
remove the nreps parameter from the f_drug_demand function
add the f_dose_observed function and incorporate it in the f_drug_demand function
modify the condition "Vi + Ti <= D(i)" to "Vi + Ti < D(i)" in the f_dose_ongoing_cpp and f_dose_new_cpp functions
change "as.numeric(exp(attr(a$apVar, "Pars")))" to "exp(as.numeric(attr(a$apVar, "Pars")))" in the f_fit_di function to avoid the error for non-numeric argument to mathematical function
simplify the condition for common_time_model to "length(unique(target_days)) == 1"
add dosing_summary_t0 to the output of the f_drug_demand function
replace mutate and slice(n()) with summarise in the f_dose_observed and f_dose_draw functions to improve efficiency
plot gap time t0 instead of t0 + 1 in the f_fit_t0 function
plot the rounded value of di based on probability calculations alongside the observed value of di in the f_fit_di function
replace the residual plot with the fitted gap time bar chart in the f_fit_ti function
redefine row_id for vf1 if common_time_model is true in the f_dispensing_models function
use df and visitview to derive treatment_by_drug_df for real-time drug demand forecasting
update the examples of the f_fit_t0, f_fit_ki, and f_fit_ti functions
add trialsdt and cutoffdt to the output of the f_dose_observed and f_drug_demand functions
move the arrange operation of dosing_subject_newi out of the f_dose_draw_1 function into the f_dose_draw function to improve efficiency
combine the two summarise operation of dosing_summary_newi in the f_dose_draw_1 function to improve efficiency
replace the zero-inflated negative binomial distribution with the negative binomial distribution in the f_fit_ki function to avoid convergence issues
print cum_dispense_plot if showplot is TRUE in the f_dose_observed function
add colors = "Set2" to cum_dispense_plot, bar_t0_plot, bar_ti_plot, and bar_di_plot in the f_dose_observed function
remove the custom legend of cum_dispense_plot in the f_dose_observed and f_drug_demand functions
add parameter l to the f_dose_draw_1 function to improve efficiency
add structure and more details to the function parameters and output descriptions
combine the dosing_subject_t and dosing_summary_t steps in the f_dose_observed function to improve efficiency
replace target_days with dosing_schedule_df in the argument for the f_dispensing_models function
drop the creation of the status variable and use table instead of survfit for observed data summary in the f_fit_t0 function
add the handling of one observation case in the f_fit_ti function
remove the creation of the unames1 variable in the f_dose_draw function