# Configuration file for simulator for treatment and outcome
# ==============================================================================

# SIMULATION
# ==============================================================================
# Type of simulation
simulation_type: "TY"

# Higher values correspond to stronger influence of predictive features
predictive_scale: 1

# Higher values correspond to stronger influence of prognostic features
prognostic_scale: 1

# Controls the overlap - higher scaling - less overlap
propensity_scale: 100

# 0 corresponds to completely balanced treatment assignment, 1 leads to strong imbalance, ->inf leads to complete imbalance
unbalancedness_exp: 0

# 0 correponds to no nonlinearity, 1 to only nonlinearity
nonlinearity_scale: 0

# Settings regarding important features
num_pred_features: 10 #for us, 10 for acic, 8 for twins
num_prog_features: 10
num_select_features: 10
feature_type_overlap: "sel_none" # "sel_none", "sel_pred", "sel_prog"
treatment_feature_overlap: True

# Propensity Settings
propensity_type: "toy" # "none_prog", "none_pred", "pred" 
alpha: 0
enforce_balancedness: True

random_feature_selection: False # Order the dataset in such way that they follow: 
# prog (# = num_prog_features), pred (# = num_T*num_pred_features), select (# = num_T*num_select_feautures) 
nonlinearity_selection_type: "random"

# Dimensionality of treatments and outcome
num_T: 2
dim_Y: 1
num_binary_outcome: 0 # How many outcomes should be binary
outcome_unbalancedness_ratio: 0.5 # 0.5 means that half of the outcomes are 0 and half are 1
standardize_outcome: False # For TY -> for each treatment to get more balanced cates, for T -> over all entries such that biological mech is not lost

# Control settings
include_control: False

# Noise settings
noise: True
noise_std: 0.05
# ==============================================================================