Certara.RDarwin (version 1.1.1)

pyDarwinOptionsPenalty: Create pyDarwin Penalty Options

Description

Generates a list of penalty parameters to be used in pyDarwin create_pyDarwinOptions function.

Usage

pyDarwinOptionsPenalty(
  theta = 10,
  omega = 10,
  sigma = 10,
  convergence = 100,
  covariance = 100,
  correlation = 100,
  condition_number = 100,
  non_influential_tokens = 1e-05
)

Value

A list of penalty options in pyDarwin optimization process.

Arguments

theta

numeric: Penalty added to fitness/reward for each estimated THETA. A value of 3.84 corresponds to a hypothesis test with 1 df and p < 0.05 (for nested models), and a value of 2 for 1 df corresponds to the Akaike information criterion. Default: 10

omega

numeric: Penalty added to fitness/reward for each estimated OMEGA element. Default: 10

sigma

numeric: Penalty added to fitness/reward for each estimated SIGMA element. Default: 10

convergence

numeric: Penalty added to fitness/reward for failing to converge. Default: 100

covariance

numeric: Penalty added to fitness/reward for failing the covariance step (real number). If a successful covariance step is important, this can be set to a large value (e.g., 100), otherwise, set to 0. Default: 100

correlation

numeric: Penalty added to fitness/reward if any off-diagonal element of the correlation matrix of estimates has an absolute value > 0.95 (real number). This penalty will be added if the covariance step fails or is not requested. Default: 100

condition_number

numeric: Penalty added if the covariance step fails or is not requested, e.g., PRINT=E is not included in $COV. Additionally, if the covariance is successful and the condition number of the covariance matrix is > 1000, then this penalty is added to the fitness/reward. Default: 100

non_influential_tokens

numeric: Penalty added to fitness/reward if any tokens do not influence the control file (relevant for nested tokens). Should be very small (e.g., 0.0001), as the purpose is only for the model with non-influential tokens to be slightly worse than the same model without the non-influential token(s) to break a tie. Default: 0.00001

Examples

# Create penalty options with default values
penalty_options <- pyDarwinOptionsPenalty()
# Create penalty options with custom values
penalty_options_custom <-
  pyDarwinOptionsPenalty(theta = 3.84,
                         omega = 8,
                         sigma = 6,
                         convergence = 50,
                         covariance = 80,
                         correlation = 60,
                         condition_number = 70,
                         non_influential_tokens = 0.0001)