openpkflow.validation

Utility functions for cross-checking openpkflow output against reference values.

Public API

Symbol	Type	Description
pct_bias(observed, reference)	function	Percent bias: (obs - ref) / ref * 100
rmse(observed, reference)	function	Root mean squared error
within_pct(observed, reference, pct)	function	True if |pct_bias| <= pct

Example

from openpkflow.validation import pct_bias, within_pct

# Check that NCA-recovered CL is within 5% of true CL
true_cl = 5.0
recovered_cl = 5.12
print(pct_bias(recovered_cl, true_cl))    # 2.4
print(within_pct(recovered_cl, true_cl, pct=5.0))  # True

Validation strategy

OpenPKFlow combines independent external comparators with analytical and internal checks. Examples include PKNCA, NonCompart, Phoenix public outputs, PowerTOST, bootf2, and R stats::nls, alongside exact closed-form PK profiles and hand-checkable degenerate cases.

See the validation matrix for comparator scope and test locations, and VALIDATION.md at the repository root for the detailed test-to-reference map. Comparator agreement guards against calculation drift; it does not by itself make OpenPKFlow a validated regulated system.