el-meyer
commented
3 years ago set.seed(1) ocs <- trial_ocs( n_int = n_int, n_fin = n_fin, random_type = random_type, rr_comb = rr_comb, rr_mono = rr_mono, rr_back = rr_back, rr_plac = rr_plac, rr_transform = rr_transform, random = random, prob_comb_rr = prob_comb_rr, prob_mono_rr = prob_mono_rr, prob_back_rr = prob_back_rr, prob_plac_rr = prob_plac_rr, stage_data = stage_data, cohort_random = cohort_random, cohorts_max = cohorts_max, sr_drugs_pos = sr_drugs_pos, target_rr = target_rr, sharing_type = sharing_type, sr_first_pos = sr_first_pos, safety_prob = safety_prob, Bayes_Sup = Bayes_Sup, prob_rr_transform = prob_rr_transform, cohort_offset = cohort_offset, trial_struc = trial_struc, iter = 50, coresnum = 1, save = FALSE, ret_list = TRUE, ret_trials = TRUE, ) set.seed(1) ocs1 <- trial_ocs( n_int = n_int, n_fin = n_fin, random_type = random_type, rr_comb = rr_comb, rr_mono = rr_mono, rr_back = rr_back, rr_plac = rr_plac, rr_transform = rr_transform, random = random, prob_comb_rr = prob_comb_rr, prob_mono_rr = prob_mono_rr, prob_back_rr = prob_back_rr, prob_plac_rr = prob_plac_rr, stage_data = stage_data, cohort_random = cohort_random, cohorts_max = cohorts_max, sr_drugs_pos = sr_drugs_pos, target_rr = target_rr, sharing_type = sharing_type, sr_first_pos = sr_first_pos, safety_prob = safety_prob, Bayes_Sup = Bayes_Sup, prob_rr_transform = prob_rr_transform, cohort_offset = cohort_offset, trial_struc = trial_struc, iter = 50, coresnum = 1, save = FALSE, ret_list = TRUE, ret_trials = TRUE, ) all.equal(ocs, ocs1)
Regarding parallelization it would seem that the implementation is rather rigid in only allowing a specific type and also it seems that parallel random number generation is not suitably implemented. See example code attached.
Also think about missing value generation.