An Optimized DEC Prediction Model Was Not the Same Test as an Officer’s Drug Opinion
A statistical model optimized from DEC observations is not the same assessment as an officer’s actual drug opinion. A laboratory study reported results for both, and their distinction matters when evaluating claims about identifying marijuana-related effects.
Two procedures used the same experimental setting
Heishman and colleagues studied 18 volunteers in nine randomized, double-blind sessions involving ethanol, cocaine, marijuana and placebo conditions. They examined a broad set of evaluation observations and compared officers’ judgments with toxicology.
The report also developed optimized prediction models using selected combinations of variables. Those models involved 17 to 28 variables. A fitted combination can use information differently from the decision process an examiner actually applied.
Do not transfer model performance to officers
The reported consistency of officers’ non-ethanol drug opinions with toxicology was 44%. That figure covered the study’s non-ethanol conditions; it was not a cannabis-specific impairment sensitivity estimate.
An optimized model’s result cannot replace that observed judgment result unless the model is actually the procedure under evaluation. Claims about deployment additionally need evidence that the specified model works on appropriate data beyond the observations used to select it.
Ask which rule produced the conclusion
A review should identify whether a quoted result belongs to a fitted equation, a predefined scoring rule or an examiner’s opinion. Those procedures can share inputs while producing different classifications and errors.
Toxicology agreement also addresses drug exposure rather than directly measuring driving ability. This controlled study helps separate these methodological questions; it does not supply a universal numerical accuracy rate for every contemporary marijuana examination.
Related: How the reference specimen changes a cannabis accuracy claim.
Research cited
Heishman SJ; Singleton EG; Crouch DJ; Journal of Analytical Toxicology; 1996. Laboratory Validation Study of Drug Evaluation and Classification Program: Ethanol, Cocaine, and Marijuana.
