1 Introduction
This document describes the FACTS XML file (“.facts”) file format for FACTS Enrichment Design (ED).
It is intended for expert FACTS users who wish to create modified versions of “.facts” files using scripts. A full and deep understanding of FATCS ED is assumed.
1.1 IDs
There are number of parameters that IDs, such as “guid”, “SetId”, … these are Microsoft “unique” identifiers, there should be no need to change these. It does not matter if the same Ids are used in successive “.facts” files, for instance if creating a range of sub-versions of a single original file.
1.2 Defaults
FACTS stores a super-set of the FACTS parameters in the “.facts” file. There are so many parameters in the FACTS file that we cannot check the behavior of FACTS if some of these are missing, so our advice is to include *all* the parameters even if not required. The only parameters that it is clear can be omitted are those in the optional parameter sets: BaselineParameterSet and LongitudinalParameterSet that are not required if the respective features are not enabled.
1.3 Parameter Order
FACTS processes the parameters in the “.facts” file in the order in which the appear in the file. Some parameters depend on other parameters to have been already defined, so it is recommended that the order given here (the order FACTS itself uses when writing out the parameters) is retained. For many parameters changing their order should not be a problem, but we are not going to define, or guarantee, which those are.
2 The values in a “.facts” file
2.1 Numbers of items
ACS: The number of studies used for the hierarchical prior for the active comparator
BS: The number of Bayesian Augment Control (BAC) studies
CHRS: The number of control hazard rates segments in TTE
CS: The number of studies used for the hierarchical prior for the control
D: The number of doses (or treatment arms) in the design
DRS: The number of dropout rate segments in TTE
G: The number of groups in the design
I: The number of interims (when defined by information)
LM: The number of longitudinal models
R: The number of regions in an accrual profile
TCM: The number of intervals (in weeks) for the Hazard model control rates
V: The number of patient visits
Var: The number of variants
LM*V: This is the value used when a different prior is used at different visits. The values are stored as lists of lists (nested <item>s). The outer dimension is the number of LM and the inner dimension the number of visits. If the priors are common across all models, just the first lists of values for visits is used, the others are present but ignored.
2.2 Common structures
2.2.1 Gamma prior
The parameters are stored in a <gamma> value as <mu> – the central value of the prior, and <n> - the weight of the prior in equivalent number of observations. E.g.:
<gamma>
<mu>10</mu>
<n>1</n>
</gamma>2.2.2 Normal prior
The parameters are stored in a <menasdpair> value as <mean> - the mean of the prior and <sd> - the standard deviation. E.g.:
<meansdpair>
<mean>0</mean>
<sd>10</sd>
</meansdpair>2.2.3 Group
The parameters stored in a <group> are
<name>: the name of the group
<cap>: flag – 0=group is not capped, 1=group is capped
<size>: the group size
E.g.:
<group>
<name>Group 1</name>
<cap>1</cap>
<size>50</size>
</group>
2.2.4 Total cap
The parameters stored in a <total_cap> is a <selected_value> object with the following data:
<selected>: flag – 0=do not apply overall cap to study, 1=apply overall cap to study
<value>: the max sample size if <selected> flag is 1.
E.g.:
<selectedvalue>
<selected>0</selected>
<value>300</value>
</selectedvalue>
2.2.5 BAC Study
The parameters stored in a <bac> are
<type>: the FACTS endpoint type (1=continuous, 2=dichotomous, 3=time to event)
<name>: the name of the BAC study
<subjects>: the number of subjects in the BAC study
<response>: the response value for the BAC study
<sd>: the SD of the BAC study
<num_responses>: the number of responses for the BAC study (Only used for TTE)
E.g.:
<bac>
<type>1</type>
<name>Study 3</name>
<subjects>300</subjects>
<response>0</response>
<sd>0.5</sd>
<num_response>0</num_response>
</bac>
2.2.6 Interim Criteria
The parameters stored in an <interim_criteria> are
interim_number: the number of the interim the criteria apply to, the number 32767 indicates the final success/futility criteria
<success_criteria>: List of G+1 <posterior_cut_list_tag> and G+1 < phase_three_cut_list_tag >, for each group and across groups. See below for details.
<futility_criteria> List of G+1 <posterior_cut_list_tag> and G+1 < phase_three_cut_list_tag >, for each group and across groups. See below for details.
2.2.7 Success / Futility Criteria
The parameters stored in a <success_criteria> or <futility_criteria> are:
posterior_cut_list_tag:
- posterior_cut_tag: Threshold for the posterior probabilities of CSD.
phase_three_cut_list_tag:
phase_three_cut_tag: Threshold for the phase III success probabilities.
2.3 The values
| Tags | Parameter | Number of items | Default Value | Notes |
|---|---|---|---|---|
| <facts | version | 1 | the FACTS GUI version number | FACTS will not open “.facts” files with a higher version number. This test has/will be relaxed to only check the first 2 numbers in the version number. |
| guid | 1 | see general comment on id’s in section 1 | ||
| name | 1 | the name of the “.facts” file | Don’t make it different from the file name, bad things may happen. | |
| host | 1 | |||
| <project-notes | CDATA | 1 | “” | All text from the “Notes” text box |
| </project-notes> | ||||
| <parameterSets | type | 1 | AIPFParameterSet | used for all FACTS Enrichment Design parameters |
| <parameterSet | name | 1 | “aipf” | |
| SetID | 1 | see general comment on id’s in section 1 | ||
| <property> | type | 1 | 1-3 | The Endpoint type: 1= Continuous, 2 = Dichotomous, 3 = TTE |
| <property> | is_adaptive | 1 | 0 | This is the “Enable adaptive features” option on the Study Info tab. If the flag=1, then the Adaptive box is checked and adaptive options displayed. It needs to be set if adaptive features (interims, allocation) are being used. |
| <property> | use_longitudinal | 1 | 1 | Flag – 0=Don’t use longitudinal modelling, 1=Use longitudinal modelling |
| <property> | enable_freq | 1 | 1 | Flag – 0=Don’t use frequentist analysis, 1=Use frequentist analysis |
| <property> | include_control | 1 | 1 | Flag – 0=Compare to objective control, 1=Include control arm |
| <property> | use_stable_state | 1 | 0 | An option for the Dichotomous endpoint. Flag – 0=Do not use restricted Markov model, 1=Use restricted Markov model |
| <property> | high_is_good | 1 | 1 | This flag corresponds to the “Higher/Lower response is subject improvement” selection on the Study Info tab. 0=Lower, 1=Higher. If this is a Dichotomous endpoint design it corresponds to the equivalent “Response indicates positive/negative outcome” selection on the Study Info tab. 0=negative, 1=positive |
| <property> | is_noninferiority | 1 | 0 | Flag defining the trial type: 0=Superiority, 1=Non-inferiority. |
| <property> | groups | G | This represents the list of groups in the design, the value of each is a group (as listed in the “2.2.3 Group”) | |
| <property> | total_cap | 1 | This represents the data for the overall cap to the study (see “2.2.4 Total cap”) | |
| <property> | include_baseline | 1 | 0 | An option for the Continuous endpoint. A flag to indicate whether including simulation of baseline: 0=Don’t simulate baseline, 1= Simulate baseline |
| <property> | is_change_baseline | 1 | 0 | An option for the Continuous endpoint, if simulating baseline. A flag to indicate whether response analyzed is absolute or change from baseline: 0=Change from baseline, 1= Final endpoint value |
| <property> | final_stable_success | 1 | 0 | An option for the Dichotomous endpoint. If use_special_long=1 AND special_long_feature=1, this flag indicates whether a subject whose final state is “stable” will be counted as a Responder or Failure, 0=Failure, 1=Responder |
| <property> | use_min_follow | 1 | 0 | An option for the TTE endpoint. A flag to indicate how follow-up is performed: 0=Use max follow-up per subject, 1=Use follow-up after full enrollment |
| <property> | min_follow | 1 | 0 | Follow-up after full enrollment (wks) |
| <property> | max_follow | 1 | 0 | Max follow-up per subject (wks) |
| <property> | has_variants | 1 | 0 | A flag to indicate if there are variants. |
| <property> | num_variants | 1 | 5 | if has_variants=1 this value is the number of variants |
| <property> | variant_names | Var | “Var 1”, “Var 2”, … | The names of the “variants” to be included in the scenario names to indicate which variant was being simulated. |
| <property> | var_group_caps | Var*G | “50, 50, …”, “50, 50, …”, … | The group caps for each of the “variants”, represents as a list of lists. Each entry is a list of caps for each of the groups of the study for a given variant. |
| <property> | success_delta | G + 1 | 0.5, 0.5, … for dich 0.2, 0.2, … for TTE 1, 1, …for cont |
This represents the list of target mean differences for success (Success CSD) for each group of the study, and across groups. See the Group Info tab in FACTS. |
| <property> | futility_delta | G + 1 | 0.3, 0.3, … for dich 0, 0, … for TTE 0.5, 0.5, …for cont |
This represents the list of target mean differences for futility (Futility CSD) for each group of the study, and across groups. See the Group Info tab in FACTS. |
| <property> | p3_pred | 1 | 0 | Flag for enabling phase III prediction criteria |
| <property> | p3_non_inf | 1 | 0 | Flag defining the phase III trial type: 0=Superiority, 1=Non-inferiority. |
| <property> | p3_arm_size | G + 1 | 250, 250, … | This represents the phase III number of subjects per arm for each group of the study, and across groups. See the Group Info tab in FACTS. |
| <property> | p3_alpha | G + 1 | 0.025, 0.025, … | This represents the phase III one-sided alpha for each group of the study, and across groups. See the Group Info tab in FACTS. |
| <property> | p3_csd | G + 1 | 0, 0, … | This represents the list of superiority or non-inferiority margins (Phase III CSD) for each group of the study, and across groups. See the Group Info tab in FACTS. |
| <property> | study_weeks | V | Only used for continuous and dichotomous endpoints. If is_longitudinal_modeled=0 this is an empty <item /> list, if is_longitudinal_modeled=1, this is a list of items, one per visit, the value of each is the week of the visit. (As listed in the “Schedule of Post-Baseline Visits) | |
| <property> | final_visit | 1 | 1 | Only used for continuous and dichotomous endpoints. This is the week of the last visit, if is_longitudinal_modeled=1 this value should be the same as the last entry in the study_weeks list. |
| <property> | visit_names | V | Only used for continuous and dichotomous endpoints. If is_longitudinal_modeled=0 this is an empty <item /> list, if is_longitudinal_modeled=1, this is a list of items, one per visit, the value of each is the visit name. (As listed in the “Schedule of Post-Baseline Visits” on the Study Info tab) | |
| <property> | interval_censor | 1 | 0 | Flag for enabling restricting event times to visits for TTE endpoints |
| <property> | time_cut_weeks | V | Only used for TTE endpoints. This is a list of items, one per visit, the value of each is the visit week. | |
| <property> | time_cut_names | V | Only used for TTE endpoints. This is a list of items, one per visit, the value of each is the visit name. | |
| <property> | control_obj_resp | G | 0, 0, … | Only used for continuous and dichotomous endpoints. This represents a list of control model means for each group when the control selection on the Study info tab is “Compare to objective control” |
| Design > Control Model (Continuous and Dichotomous) | ||||
| <property> | hier_borrow_control | 1 | 1 | A flag to indicate how the priors for control are modelled: 0=Controls modelled separately, 1=Hierarchical model across groups |
| <property> | hier_mu | 1 | 0 | |
| <property> | hier_sd | 1 | 1 | when when hier_borrow_control=1 (Hierarchical model across groups). This represents the Control model hierarchical prior SD for mu |
| <property> | hier_gamma | <gamma> | 1, 1 | Only used when hier_borrow_control=1 (Hierarchical model across groups). Contains the values for the Control model inverse-gamma prior for ‘tau’ squared, the variance of the mean of the endpoint |
| <property> | control_prior_mean | G | 0, 0, … | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, the associated Control model value prior mean |
| <property> | control_prior_sd | G | 1, 1, … | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, the associated Control model prior SD |
| <property> | control_historical_data | G | 0, 0, … | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, a flag representing whether to use a hierarchical prior for the associated Control model |
| <property> | bac_mu | G | (0, 1), (0, 1),… | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, a <gamma> parameter representing the mean and SD of the group’s mean Mu (µ_H,g) for the associated Control model |
| <property> | bac_tau | G | (1,1), (1,1), … | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, a <gamma> parameter representing the mean and SD of the group’s variance Tau^2 (τ_H,g^2) for the associated Control model |
| <property> | bac_studies | BS | Only used when hier_borrow_control=0 (Controls modelled separately). Contains, for each group, a list of <bac> for the associated Control model | |
| <property> | gui_cluster_control | 1 | 0 | Only used when hier_borrow_control=1 (Hierarchical model across groups). Flag for enabling clustering for the Control model |
| <property> | dp_scale_param_control | 1 | 2 | Only used when gui_cluster_control=1. The clustering Dp scale parameter for the Control model |
| Design > Hazard Model (TTE) | ||||
| <property> | cox_model | 1 | 0 | Flag for enabling the Cox proportional hazard model |
| <property> | time_cuts_mod | TCM | Only used when cox_model=0. This represents the list of intervals for control rates (wks) | |
| <property> | control_rates | TCM*G | Only used when include_control=0. For any given interval of controls rates (wks), this represents the Control hazard rate for each of the groups of the study | |
| <property> | lambda_mean | G*TCM | Only used when cox_model=0 and hier_borrow_control=0 (Controls modelled separately). For any given group this represents the prior mean on the Gamma distribution for the control arm hazard rate | |
| <property> | prior_lambda_a | G*TCM | Only used when cox_model=0 and hier_borrow_control=0 (Controls modelled separately). For any given group this represents the prior weight on the Gamma distribution for the control arm hazard rate | |
| <property> | bac_studies_names | G*CS | “” | If the entry of control_historical_data associated with a given group is 1, this is the list of study names for the hierarchical prior of Control, for that given group |
| <property> | bac_study_data | G*TCM | This is a derived property that is not directly set through the GUI. Rather, it agglomerates all the BAC study data. If modifying this, ensure it is consistent in length and content with the values of bac_studies_names and time_cuts_mod | |
| <property> | prior_lambda_a_mean | TCM | 1, 1, … | Only used when cox_model=0 and hier_borrow_control=1 (Hierarchical model across groups). This represents the list of hyperprior means on the Gamma distribution for αs, for each interval |
| <property> | prior_lambda_a_alpha | TCM | 1, 1, … | Only used when cox_model=0 and hier_borrow_control=1 (Hierarchical model across groups). This represents the list of hyperprior weights on the Gamma distribution for αs, for each interval |
| <property> | prior_lambda_b_mean | TCM | 1, 1, … | Only used when cox_model=0 and hier_borrow_control=1 (Hierarchical model across groups). This represents the list of hyperprior means on the Gamma distribution for βs, for each interval |
| <property> | prior_lambda_b_alpha | TCM | 1, 1, … | Only used when cox_model=0 and hier_borrow_control=1 (Hierarchical model across groups). This represents the list of hyperprior weights on the Gamma distribution for βs, for each interval |
| Design > Treatment Model | ||||
| <property> | hier_borrow_trt | 1 | 1 | Flag for how prior for treatment are handled: 0=Groups modeled separately, 1=Hierarchical model across groups |
| <property> | trt_prior_mu | G | 0, 0, … | Only used when hier_borrow_trt=0. Contains the prior means of the treatment model for each group |
| <property> | trt_prior_sd | G | 1, 1, … | Only used when hier_borrow_trt=0. Contains the prior SD of the treatment model for each group |
| <property> | hier_trt_mu | 1 | 0 | Only used when hier_borrow_trt=1. This is the hierarchical prior mean for Mu (µ) in the treatment model |
| <property> | hier_trt_sd | 1 | 1 | Only used when hier_borrow_trt=1. This is the hierarchical prior SD for Mu (µ) in the treatment model |
| <property> | hier_trt_gamma | <gamma> | (1, 1) | Only used when hier_borrow_trt=1. Contains the values for the treatment model inverse-gamma prior for ‘tau’ squared, the variance of the mean of the endpoint |
| <property> | sigma_prior_gamma | <gamma> | (1, 1) | Only used for continuous endpoints. Contains the values for the inverse-gamma prior for ‘sigma’ squared, the estimate of the variance of the endpoint. |
| <property> | baseline_adjusted | 1 | 0 | Flag to indicate if the subjects final response is to be adjusted for baseline, 0: No adjustment, 1: Adjusted |
| <property> | baseline_beta | <meansdpair> | (0, 1) | Only used when baseline_adjusted=1. The prior for the baseline adjustment model |
| <property> | missing_data_model | 1 | 0 | This indicates the type of missing data model to use as part of the treatment model, 0: Bayesian multiple imputation from post baseline, 1: Baseline observation carried forward (BOCF), 2: Last observation carried forward (LOCF) |
| <property> | pooled_prior_mu | 1 | 0 | The prior mean for the ‘Across Group’ model |
| <property> | pooled_prior_sd | 1 | 1 | The prior SD for the ‘Across Group’ model |
| <property> | gui_cluster_trt | 1 | 0 | Flag for enabling clustering for the treatment model |
| <property> | dp_scale_param_trt | 1 | 2 | Only used when gui_cluster_trt=1. The clustering Dp scale parameter for the treatment model |
| Design > Longitudinal | ||||
| <property> | long_model_type | 1 | 3 | Flag that indicates the type of the longitudinal model. For a Continuous endpoint these are: 1=linear regression, 3=last observation carried forward, 4=time course hierarchical, 5=kernel density, 8=ITP For a Dichotomous endpoint: 2=Beta binomial 3=last observation carried forward, 7=Logistic regression |
| <property> | long_model_spec | 1 | 0 | This flag indicates the number of longitudinal model instances 0=Single model for all groups and arms 1=Model arms separately 2=Model groups separately 3=Model groups and arms separately 4=Model all arms separately |
| <property> | long_prior_spec | 1 | 0 | This flag indicates the number of separate priors for the longitudinal model parameters 0=Same prior across all model instances and visits 1=Specify per visit, across all model instances 2=Specify priors per model instance and visit |
| <property> | long_model_instances | LM | This represents the list of longitudinal model instance names. Its content and length depends on the value of long_model_spec. long_model_spec=0: “All groups” long_model_spec=1: “Control arms”, “Treatment arms” long_model_spec=2: “[Group name]” long_model_spec=3: “[Group name] control”, “[Group name] treatment]” |
|
| <property> | long_alpha_mu | 1 | 0 | When long_prior_spec=0, contains the values for the mean of the Normal prior for the intercept of the linear regression, models (αt) to be used for all models This is the value used when the same prior is used across all model instances and visits. |
| <property> | long_alpha_mu_full | LM*V | 0 | When long_prior_spec=1 or 2, contains the values for the mean of the Normal prior for the intercept of the linear regression, models (αt) with different values for each visit, and possibly each model. See the description of LM*V parameters before this table. |
| <property> | long_alpha_stdv | 1 | 10 | When long_prior_spec=0, contains the values for the SD of the Normal prior for the intercept of the linear regression, models (αt) to be used for all models This is the value used when the same prior is used across all model instances and visits. |
| <property> | long_alpha_sd_full | LM*V | 10 | When long_prior_spec=1 or 2, contains the values for the SD of the Normal prior for the intercept of the linear regression, models (αt) with different values for each visit, and possibly each model. See the description of LM*V parameters before this table. |
| <property> | long_beta_mu | 1 | 0.75 | When long_prior_spec=0, contains the values for the mean of the Normal prior for the slope of the linear regression, models (βt) to be used for all models This is the value used when the same prior is used across all model instances and visits. |
| <property> | long_beta_mu_full | LM*V | 0.75 | When long_prior_spec=1 or 2, contains the values for the mean of the Normal prior for the slope of the linear regression, models (βt) with different values for each visit, and possibly each model. See the description of LM*V parameters before this table. |
| <property> | long_beta_stdv | 1 | 1 | When long_prior_spec=0, contains the values for the SD of the Normal prior for the slope of the linear regression, models (βt) to be used for all models This is the value used when the same prior is used across all model instances and visits. |
| <property> | long_beta_sd_full | LM*V | 1 | When long_prior_spec=1 or 2, contains the values for the SD of the Normal prior for the slope of the linear regression, models (βt) with different values for each visit, and possibly each model. See the description of LM*V parameters before this table. |
| <property> | long_lambda | <gamma> | 0.5, 1 | Contains the values for the inverse-gamma prior for the variance of the residual error of the linear regression, time course hierarchical and ITP longitudinal models (λ2). This is the value used when the same prior is used across all model instances and visits. |
| <property> | long_lambda_list | <gamma>*LM | 0.5, 1 | Contains the values for the inverse-gamma prior for the variance of the residual error of the time course hierarchical and ITP longitudinal models (λ2). This is the value used when a different prior is used across all model instances. These models do not have separate estimates of this parameter per visit, unlike the linear regression… |
| <property> | long_lambda_full | <gamma>*LM*V | 0.5, 1 | Contains the values for the inverse-gamma prior for the variance of the residual error of the linear regression longitudinal model (λ2). See the description of LM*V parameters before this table. |
| <property> | tch_tau | <gamma>*LM | 0.5, 1 | Contains the values for the inverse-gamma prior for the variance of the inter-subject error term (δi) of the Time Course Hierarchical longitudinal model (τ2). The number of prior values depends on the number of longitudinal models, even if the same priors are used across all models. In which case the first values are used for all models. |
| <property> | tch_alpha_matrix | <meansdpair>*LM*V | 0, 2, … | Whether long_prior_spec=0 or 2, these are the priors for the αt parameter in the Time Course Hierarchical longitudinal model. αt is the log of the fraction of the final response seen at visit t. Unlike other LMs there always separate prior values for each visit. If long_prior_spec=0 only the values for the first model in the list are set. See the description of LM*V parameters before this table. |
| <property> | kernel_bandwidth | LM | 2 | This is the value of the kernel bandwidth parameter for the Kernel Density longitudinal model. The value is a list of items - one for each LM model. If long_prior_spec=0, and there are multiple models, just the first value in the list is used, even though entries for the other models are present. |
| <property> | kernel_min_subj | LM | 6 | This is the value of the “minimum number of subjects before using” parameter for the Kernel Density longitudinal model. The value is a list of items - one for each LM model. If long_prior_spec=0, and there are multiple models, just the first value in the list is used, even though entries for the other models are present. |
| <property> | kernel_sd | LM | 2 | This is the value of the “standard deviation of the prior mean” parameter for the Kernel Density longitudinal model. The value is a list of items - one for each LM model. If long_prior_spec=0, and there are multiple models, just the first value in the list is used, even though entries for the other models are present. |
| <property> | bb_alpha_mu_1 | 1 | 1 | When long_prior_spec=0, this is the single prior for the αμ1 parameter for the Beta Binomial prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | bb_alpha_mu_1_full | LM*V | 1, 1, … | When long_prior_spec=1 or 2, these are the priors for the αμ1 parameter for the Beta Binomial prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | bb_alpha_mu_0 | 1 | 1 | When long_prior_spec=0, this is the single prior for the αμ0 parameter for the Beta Binomial prior for the probability that a subject without a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | bb_alpha_mu_0_full | LM*V | 1, 1, … | When long_prior_spec=1 or 2, these are the priors for the αμ0 parameter for the Beta Binomial prior for the probability that a subject without a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | bb_beta_mu_1 | 1 | 1 | When long_prior_spec=0, this is the single prior for the βμ1 parameter for the Beta Binomial prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | bb_beta_mu_1_full | LM*V | 1, 1, … | When long_prior_spec=1 or 2, these are the priors for the βμ1 parameter for the Beta Binomial prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | bb_alpha_mu_0 | 1 | 1 | When long_prior_spec=0, this is the single prior for the αμ0 parameter for the Beta Binomial prior for the probability that a subject without a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | bb_alpha_mu_0_full | LM*V | 1, 1, … | When long_prior_spec=1 or 2, these are the priors for the αμ0 parameter for the Beta Binomial prior for the probability that a subject without a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | stable_state_alpha_0 | LM*V | 1, 1, … | Only used for dichotomous endpoints with use_stable_state=1. Whether long_prior_spec=0 or 2, these are the priors for the α0 parameter for the Dirichlet prior in the Restricted Markov Model, for the probability that stable subject will be a fail at the final visit across the different visits. Unlike other LMs there always separate prior values for each visit. If long_prior_spec=0 only the values for the first model in the list are set. See the description of LM*V parameters before this table. |
| <property> | stable_state_alpha_S | LM*V | 1, 1, … | Only used for dichotomous endpoints with use_stable_state=1. Whether long_prior_spec=0 or 2, these are the priors for the αS parameter for the Dirichlet prior in the Restricted Markov Model, for the probability that stable subject will be stable at the final visit across the different visits. Unlike other LMs there always separate prior values for each visit. If long_prior_spec=0 only the values for the first model in the list are set. See the description of LM*V parameters before this table |
| <property> | stable_state_alpha_1 | LM*V | 1, 1, … | Only used for dichotomous endpoints with use_stable_state=1. Whether long_prior_spec=0 or 2, these are the priors for the α1 parameter for the Dirichlet prior in the Restricted Markov Model, for the probability that stable subject will be responder at the final visit across the different visits. Unlike other LMs there always separate prior values for each visit. If long_prior_spec=0 only the values for the first model in the list are set. See the description of LM*V parameters before this table |
| <property> | lr_alpha_mu_1 | 1 | 0.5 | Only used for dichotomous endpoints. When long_prior_spec=0, this is the single prior for the µ1 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | lr_alpha_mu_1_full | LM*V | 0.5, 0.5, … | When long_prior_spec=1 or 2, these are the priors for the µ1 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | lr_alpha_mu_0 | 1 | 0.5 | Only used for dichotomous endpoints. When long_prior_spec=0, this is the single prior for the µ0 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | lr_alpha_mu_0_full | LM*V | 0.5, 0.5, … | Only used for dichotomous endpoints. When long_prior_spec=1 or 2, these are the priors for the µ0 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | lr_beta_mu_1 | 1 | 0.5 | Only used for dichotomous endpoints. When long_prior_spec=0, this is the single prior for the σ1 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | lr_beta_mu_1_full | LM*V | 0.5, 0.5, … | Only used for dichotomous endpoints. When long_prior_spec=1 or 2, these are the priors for the σ1 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | lr_beta_mu_0 | 1 | 0.5 | Only used for dichotomous endpoints. When long_prior_spec=0, this is the single prior for the σ0 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit – for all models and all visits. |
| <property> | lr_beta_mu_0_full | LM*V | 0.5, 0.5, … | Only used for dichotomous endpoints. When long_prior_spec=1 or 2, these are the priors for the σ0 parameter for the Logistic Regression prior for the probability that a subject with a response at a visit will have a response at the final visit across the different visits or models and visits. See the description of LM*V parameters before this table. |
| <property> | long_itp_k | <meansdpair>*LM | 1,1 | Contains the values for the Normal prior for the ITP shape parameter (k) of the ITP longitudinal model. The number of prior values depends on the number of longitudinal models, even if the same priors are used across all models. In which case the first values are used for all models. |
| <property> | long_itp_tau | <gamma>*LM | 1,1 | Contains the values for the inverse-gamma prior for the variance of the inter-subject error term (si) (subject specific random effect) of the ITP longitudinal model (τ2). The number of prior values depends on the number of longitudinal models, even if the same priors are used across all models. In which case the first values are used for all models. |
| <property> | long_itp_omega_g | <meansdpair>*LM*G*2 | 1,1 | Contains the values for the Normal prior for the mean final response of the ITP longitudinal model, for the control arm and the treatment arm, for each group. The number of prior values depends on the number of longitudinal models, even if the same priors are used across all models. In which case the first values are used for all models. |
| Design > Allocation | ||||
| <property> | group_block_size | G*2 | 6, 6 | For each group in the study, this is the list of randomization block sizes for the Control arm and the treatment arm, respectively |
| Design > Interims | ||||
| <property> | update_mode | 1 | 0 | If is_adaptive=1, this flag indicates whether interims are: 0=by time, 1=by information |
| <property> | alloc_info_type | 1 | 0 | This is the “Subject Information Defined By:” option: 0=Subjects Enrolled, 1=Complete Data at Specified Visit, 2=Opportunity to Complete at Specified Visit, 3=Events (TTE specific) |
| <property> | update_frequency | 1 | 2 | If interims are defined by time, this is the update frequency in weeks. |
| <property> | num_burnin_subjects | 1 | 1 | If interims are by time, this is the amount of information required for the first interim to occur |
| <property> | update_by_event | 1 | 0 | Deprecated legacy flag. Previously used in TTE and now replaced by alloc_info_type==3. Please ignore. |
| <property> | update_subjects | I | 1, … | If interims are defined by the amount of information, these values are the number of subjects/events enrolled / complete / opportunity to complete at each interim |
| <property> | interims_milestones_visits | I | ““,”“, | By default blank but if interims are by the amount of information this is the visit number of the information. |
| <property> | first_interim_visit | 1 | 1 | If is_adaptive=1, and alloc_info_type=1 or 2 and alloc_update_type=0, (So only used when the interims are by time and information is completed data at specified visit or opportunity to complete at specified visit). Represents the visit number text box when specifying first interim information in the context of interims by time. |
| <property> | continue_follow_success | 1 | 1 | This is the flag for whether to follow-up subjects if the trial has stopped early for success (1=follow-up, 0=don’t follow-up) |
| <property> | continue_follow_futility | 1 | 0 | This is the flag for whether to follow-up subjects if the trial has stopped early for futility (1=follow-up, 0=don’t follow-up) |
| <property> | stop_ia | 1 | 0 | If is_adaptive=1 and alloc_update_type=0, this flag controls whether to “discontinue interim analysis beyond full enrollment”, 0=don’t discontinue, 1=discontinue |
| Design > Success/Futility Criteria | ||||
| <property> | interim_stopping_criteria | This comprises a <interims_criteria> which is a list of <interim_criteria> one per interim for which criteria have been defined plus one for the final criteria. See details above. | ||
| <property> | use_futility_stopping | 1 | 0 | Applied when is_adaptive=1. Flag indicating whether to enable futility stopping criteria for interims |
| <property> | use_fut_post | 1 | 1 | Only used when use_futility_stopping=1. Flag indicating whether to enable posterior probability of CSD as futility criteria |
| <property> | use_fut_p3 | 1 | 0 | Only used when p3_pred=1 and use_futility_stopping=1. Flag indicating whether to enable phase III success probability as futility criteria |
| <property> | futility_study_or | 1 | 1 | Flag indicating how futility criteria rules get combined, 0=AND, 1=OR |
| <property> | total_subj_fut | 1 | 0 | Only used when use_futility_stopping=1. Minimum total subjects before study can be stopped for futility |
| <property> | total_event_fut | 1 | 0 | Only used for TTE endpoints and when use_futility_stopping=1. Minimum total events before study can be stopped for futility |
| <property> | futility_group | G + 1 | 0, 0, … | List of flags for enabling early stopping for futility, for each group and across groups |
| <property> | futility_study_stop | G + 1 | 0, 0, … | List of flags for enabling the study to stop for futility if any selected groups (incl. across groups) stop. A given entry can only be set to 1 if the associated entry in futility_group is set to 1 |
| <property> | futility_stop_type | G + 1 | 0, 0, … | Represents the list of futility group stopping rule types, for each group and across groups. Group stopping rule types: 0=Group only, 1=Group AND Across Groups, 2=Group OR Across Groups |
| <property> | min_subj_fut | G + 1 | 0, 0, … | Only used when use_futility_stopping=1. List of minimum subjects in a group before study can be stopped for futility, for each group and across groups |
| <property> | min_event_fut | G + 1 | 0, 0, … | Only used for TTE endpoints and when use_futility_stopping=1. List of minimum events in a group before study can be stopped for futility, for each group and across groups |
| <property> | use_success_stopping | 1 | 0 | Applied when is_adaptive=1. Flag indicating whether to enable success stopping criteria for interims |
| <property> | use_suc_post | 1 | 1 | Only used when use_success_stopping=1. Flag indicating whether to enable posterior probability of CSD as success criteria |
| <property> | use_suc_p3 | 1 | 0 | Only used when p3_pred=1 and use_success_stopping=1. Flag indicating whether to enable phase III success probability as success criteria |
| <property> | success_study_or | 1 | 0 | Flag indicating how success criteria rules get combined, 0=AND, 1=OR |
| <property> | total_subj_succ | 1 | 0 | Only used when use_success_stopping=1. Minimum total subjects before study can be stopped for success |
| <property> | total_event_succ | 1 | 0 | Only used for TTE endpoints and when use_success_stopping=1. Minimum total events before study can be stopped for success |
| <property> | success_group | G + 1 | 0, 0, … | List of flags for enabling early stopping for success, for each group and across groups |
| <property> | success_study_stop | G + 1 | 0, 0, … | List of flags for enabling the study to stop for success if any selected groups (incl. across groups) stop. A given entry can only be set to 1 if the associated entry in success_group is set to 1 |
| <property> | success_stop_type | G + 1 | 0, 0, … | Represents the list of success group stopping rule types, for each group and across groups. Group stopping rule types: 0=Group only, 1=Group AND Across Groups, 2=Group OR Across Groups |
| <property> | min_subj_succ | G + 1 | 0, 0, … | Only used when use_success_stopping=1. List of minimum subjects in a group before study can be stopped for success, for each group and across groups |
| <property> | min_event_succ | G + 1 | 0, 0, … | Only used for TTE endpoints and when use_success_stopping=1. List of minimum events in a group before study can be stopped for success, for each group and across groups |
| <property> | use_fut_post_eval | 1 | 1 | Flag indicating whether to enable posterior probability of CSD as futility criteria for the final evaluation |
| <property> | use_fut_p3_eval | 1 | 0 | Only used when p3_pred=1. Flag indicating whether to enable phase III success probability as futility criteria for the final evaluation |
| <property> | futility_study_eval_or | 1 | 0 | Flag indicating how futility criteria rules get combined for the final evaluation, 0=AND, 1=OR |
| <property> | futility_cuts_fin | G + 1 | 0.4, 0.4, … | List of futility criteria thresholds for the posterior probabilities of CSD for the final evaluation, for each group and across groups |
| <property> | futility_cuts_p3_fin | G + 1 | 0.4, 0.4, … | List of futility criteria thresholds for the phase III success probabilities for the final evaluation, for each group and across groups |
| <property> | futility_study_eval | G + 1 | 0, 0, … | List of flags for enabling the study to be futile if all selected groups (incl. across groups) are futile; for the final evaluation |
| <property> | futility_eval_type | G + 1 | 0, 0, … | Represents the list of futility group evaluation rule types, for each group and across groups. Group evaluation rule types: 0=Group only, 1=Group AND Across Groups, 2=Group OR Across Groups |
| <property> | use_suc_post_eval | 1 | 1 | Flag indicating whether to enable posterior probability of CSD as success criteria for the final evaluation |
| <property> | use_suc_p3_eval | 1 | 0 | Only used when p3_pred=1. Flag indicating whether to enable phase III success probability as success criteria for the final evaluation |
| <property> | success_study_eval_or | 1 | 0 | Flag indicating how success criteria rules get combined for the final evaluation, 0=AND, 1=OR |
| <property> | success_cuts_fin | G + 1 | 0.4, 0.4, … | List of success criteria thresholds for the posterior probabilities of CSD for the final evaluation, for each group and across groups |
| <property> | success_cuts_p3_fin | G + 1 | 0.4, 0.4, … | List of success criteria thresholds for the phase III success probabilities for the final evaluation, for each group and across groups |
| <property> | success_study_eval | G + 1 | 0, 0, … | List of flags for enabling the study to be success if all selected groups (incl. across groups) are success; for the final evaluation |
| <property> | success_eval_type | G + 1 | 0, 0, … | Represents the list of success group evaluation rule types, for each group and across groups. Group evaluation rule types: 0=Group only, 1=Group AND Across Groups, 2=Group OR Across Groups |
| Design > Frequentist Analysis | ||||
| <property> | overall_alpha | 1 | 0.05 | Value of the ‘across groups’ two-sided alpha level |
| <property> | interact_alpha | 1 | 0.05 | Value of the ‘interaction term’ two-sided alpha level |
| <property> | group_alpha | G | 0.05, 0,05, … | List of two-sided alpha levels per group |
| </parameterSet> | ||||
| </parameterSets> | ||||
| Execution > Dropout Rate | ||||
| <parameterSets | type | 1 | AIPFDropoutParamSet | The collection of dropout profiles |
| <parameterSet | name | <Profile name> | The name of the dropout profile | |
| SetID | see general comment on id’s in section 1 | |||
| <property> | mode | 1 | 0 | flag used to indicate the type of dropout profile 0: no dropouts 1: dropouts per visit in Cont and Dich / dropouts per group, arm and visit in TTE 2: dropouts per arm and visit (Cont and Dich only) 3: dropouts per group, arm and visit in Cont and Dich / dropouts per group and arm in TTE 4: dropouts per group, arm and segment (TTE only) |
| <property> | rate | V | 0.05, 0.05, … | Only used when mode=1. This represent the list of dropout probabilities for each visit |
| <property> | treat_rate | V | 0.05, 0.50, … | Only used when mode=2. This represent the list of treatment arm dropout probabilities for each visit |
| <property> | control_rate | V | 0.05, 0.05, … | Only used when mode=2. This represent the list of Control arm dropout probabilities for each visit |
| <property> | treat_group_rate | G*V | 0.05, 0.05, … | Only used when mode=3. For each group, this represent the list of treatment arm dropout probabilities for each visit |
| <property> | control_group_rate | G*V | 0.05, 0.05, … | Only used when mode=3. For each group, this represent the list of Control arm dropout probabilities for each visit |
| <property> | treat_group_hazard | G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=3. This represent the list of treatment arm dropout hazard rates for each group |
| <property> | control_group_hazard | G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=3. This represent the list of Control arm dropout hazard rates for each group |
| <property> | treat_visit_rate | V*G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=1. For each visit, this represent the list of treatment arm dropout probabilities for each group |
| <property> | control_visit_rate | V*G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=1. For each visit, this represent the list of Control arm dropout probabilities for each group |
| <property> | segment_starts_drop | DRS | 0.0, 0.0, … | Only used for TTE endpoints and when mode=4. This represents a list of segment starting points for specifying dropout rates |
| <property> | treat_segment_hazard | DRS*G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=4. For each segment, this represent the list of treatment arm dropout hazard rates for each group |
| <property> | control_segment_hazard | DRS*G | 0.05, 0.05, … | Only used for TTE endpoints and when mode=4. For each segment, this represent the list of Control arm dropout hazard rates for each group |
| </parameterSet> | ||||
| </parameterSets> | ||||
| VSR > Control Hazard Rates (TTE) | ||||
| <parameterSets | type | 1 | AIPFControlHazardRateParameterSet | The collection of VSR control hazard rates profiles |
| <parameterSet | name | <Profile name> | The name of the control hazard rates profile | |
| SetID | see general comment on id’s in section 1 | |||
| <property> | time_cuts_sim | CHRS | 0.0, 0.0, … | Only used for TTE endpoints. This represents a list of segment starting points for specifying VSR Control hazard rates |
| <property> | true_lambda_arb | CHRS *G | 1, 1, … | Only used for TTE. For each segment, this represent the list of number of events that have occurred for each group |
| <property> | external_vsr_mean_hr | 1 | 0 | Flag for enabling the loading of external Control hazard rate means |
| <property> | external_mean_vsr_file_hr | 1 | null | Name of the file containing the external Control hazard rate means |
| </parameterSet> | ||||
| </parameterSets> | ||||
| VSR > Group Response | ||||
| <parameterSets | type | 1 | AIPFGroupResponseSet | The collection of VSR group response profiles |
| <parameterSet | name | <Profile name> | The name of the group response profile | |
| SetID | see general comment on id’s in section 1 | |||
| <property> | vsr_treatment_response | G | 0, 0, … for cont 0.5, 0.5, … for dich 0.9, 0.9, … for TTE |
The list of true responses for the treatment arm for each group |
| <property> | vsr_control_response | G | 0, 0, … for cont 0.2, 0.2, … for dich 1, 1, … for TTE |
Only used for continuous or dichotomous endpoints. The list of true responses for the Control arm for each group |
| <property> | sigma_spec | 1 | 0 | Only used for continuous endpoints. This represents the option for how the VSR SD response should be specified: 0=single SD response 1=SD response per arm 3=SD response per group and arm |
| <property> | final_sigma | 1 | 12 | Only used when sigma_spec=0. This is the value of the single SD response |
| <property> | final_treat_sigma | 1 | 12 | Only used when sigma_spec=1. This is the value of the treatment arm SD response |
| <property> | final_control_sigma | 1 | 12 | Only used when sigma_spec=1. This is the value of the Control arm SD response |
| <property> | treat_sigma | G | 10, 10, … | Only used when sigma_spec=3. This is the list of values of the treatment arm SD response for each group |
| <property> | control_sigma | G | 10, 10, … | Only used when sigma_spec=3. This is the list of values of the Control arm SD response for each group |
| <property> | external_vsr_mean | 1 | 0 | Flag for enabling the loading of external Group response means |
| <property> | external_mean_vsr_file | 1 | null | Name of the file containing the external Group response means |
| <property> | should_succeed | G | 0, 0, … | List of flags indicating whether each group in the study should succeed |
| </parameterSet> | ||||
| </parameterSets> | ||||
| VSR > Longitudinal | ||||
| <parameterSets | type | 1 | AipfLongitudinalParamSet | The collection of VSR longitudinal profiles |
| <parameterSet | name | <Profile name> | The name of the longitudinal profile | |
| SetID | see general comment on id’s in section 1 | |||
| <property> | sim_method | 1 | 1 if cont 5 if dich |
The simulation method: 1: Correlation 2: Hierarchical 5: Enter / Retain (Dichotomous only) 6: ITP 8: Hierarchical MMRM |
| <property> | set_per_group | 1 | 1 | This flag indicates the VSR longitudinal parameters related to the study groups. 0=Define parameters for all groups, 1=Define parameters for each group |
| <property> | vsr_instances | G | “Group 1”, “Group 2”, … if set_per_group=0; “All Groups” if set_per_group=1 |
This is the list of VSR instance names associated with the longitudinal profile, for each group |
| <property> | correlation | V*G | 0.8, 0.8, … | Only used when sim_method=1. For each visit, this represent the list of visit correlation probabilities, for each group |
| <property> | fraction | V*G | 1, 1, … | For each visit, this represent the list of response fractions, for each group |
| <property> | fraction_sigma | V*G | 1, 1, … | For each visit, this represent the list of fractions of the final variance (φ^2), for each group |
| <property> | omega | 1 | 0.5 | Only used when set_per_group=0. This is the fraction of overall variance due to intersubject variance, for the Control arm |
| <property> | omega_trt | 1 | 0.5 | Only used when set_per_group=0. This is the fraction of overall variance due to intersubject variance, for the treatment arm |
| <property> | omega_per_group | 2*G | 0.5 | Only used when set_per_group=1. For the Control arm and the treatment arm respectively, this is the list of fractions of overall variance due to intersubject variance, for each group |
| <property> | omega_per_group | 2*G | 0.5 | Only used when set_per_group=1. For the Control arm and the treatment arm respectively, this is the list of fractions of overall variance due to intersubject variance, for each group |
| <property> | true_retain_per_visit | V | 0.9, 0.9, … | Only used for dichotomous endpoints and when sim_method=5. This represents the list of transition probabilities r_t in the Enter / Retain VSR model for each visit |
| <property> | true_enter_per_visit | V | 0.2, 0.2, … | Only used for dichotomous endpoints and when sim_method=5. This represents the list of transition probabilities q_t in the Enter / Retain VSR model for each visit |
| <property> | itp_k_ctrl | 1 | 1 | Only used when set_per_group=0. This is the value of the Control arm ITP shape parameter k of the ITP longitudinal model for all groups. |
| <property> | itp_k_trt | 1 | 1 | Only used when set_per_group=0. This is the value of the treatment arm ITP shape parameter k of the ITP longitudinal model for all groups. |
| <property> | itp_k | 2*G | 1, 1, … | Only used when set_per_group=1. For each of the Control arm and the treatment arm respectively, this contains the values of the ITP shape parameter k of the ITP longitudinal model for each group |
| </parameterSet> | ||||
| </parameterSets> | ||||
| Accrual Rate | ||||
| <parameterSets | type | 1 | GroupRecruitmentParamSet | The collection of accrual profiles |
| <parameterSet | name | <Profile name> | The name of the accrual profile | |
| SetID | see general comment on id’s in section 1 | |||
| <property> | is_stage_two_data | 1 | 0 | Unused in Enrichment designs. Please ignore |
| <property> | location_names | R=number of profiles in accrual profile | <Region name> | The name of the regions – with an entry (<item>) for each region. All properties in this profile should have the same number of items. |
| <property> | recruit_rates | R | 5 | The peak mean accrual rate per region |
| <property> | start_dates | R | 0 | The start date (in weeks from the start of the trial) of accrual per region |
| <property> | ramp_up | R | 0 | A flag per region indicating if there is an initial ramp up stage (0= no ramp up, 1=ramp up) |
| <property> | peak_start | R | Include but leave blank if no ramp up. A per region value of the end (in weeks from the start of the trial) of ramp up in that region | |
| <property> | ramp_down | R | 0 | A flag per region indicating if there is a ramp down stage (0= no ramp down, 1=ramp down) |
| <property> | start_day | R | Include but leave blank if no ramp down. A per region value of the start (in weeks – despite the parameter name - from the start of the trial) of ramp down in that region | |
| <property> | last_day | R | Include but leave blank if no ramp down. A per region value of the end (in weeks – despite the parameter name - from the start of the trial) of ramp down in that region | |
| <property> | common_accrual_profile | R | 1 | Unused in Enrichment designs. Please ignore |
| <property> | pop_ratio | G | 1, 1, … | The list of group accrual relative group proportions, for each group |
| <property> | start_week | G | 0, 0, … | The list of group accrual start weeks, for each group |
| </parameterSet> | ||||
| </parameterSets> | ||||
| Simulation Settings | ||||
| <simulation-settings | ||||
| <property> | num_sims | 1 | 5 | The number of simulations to run |
| <property> | burn | 1 | 1000 | The number of iterations of the MCMC loop that are initially ignored before sampling starts |
| <property> | length | 1 | 2500 | The number of iterations of the MCMC loop used to estimate the parameters |
| <property> | rv_seed_1 | 1 | 3500 | The first part (32 bits) of the random number seed for all simulations. |
| <property> | rv_seed_2 | 1 | 1 | The second part (32 bits) of the random number seed for all simulations. |
| <property> | num_weeks_out | 1 | 100 | The number of simulations which will have their interim analysis results written out to file. |
| <property> | num_freq_weeks_out | 1 | 1 | The number of simulations which will have their interim frequentist analysis results written out to file. |
| <property> | num_subj_out | 1 | 1 | The number of simulations which will have their subject data written out to file. |
| <property> | num_mcmc_files | 1 | 0 | The number of simulations which will have their MCMC samples written out to file. |
| <property> | mcmc_thin | 1 | 1 | This is the factor by which the MCMC samples are thinned *when writing out the MCMC file*, it does not thin the samples used within FACTS. There is no option to thin the sampling within FACTS. |
| <property> | mcmc_per_imputation | 1 | 1 | The number of MCMC samples between each imputation of missing data using the longitudinal model |
| <property> | is_local | 1 | 1 | Flag specifying where FACTS simulations are performed. 0=grid, 1=locally |
| </simulation-settings> | ||||
| </facts> |