Monday, June 16, 2014

Junk Science, Daubert and the 2 Zone Model

You may have heard about the Daubert legal ruling.  It is an attempt to keep junk science out of the courtroom.  The Daubert standard governs the admissibility of expert witness testimony during all U.S. federal legal proceedings and over half of state proceedings. It allows for the legal challenge of any expert witness testimony. In essence, the expert scientific witness has the burden of proof relative to the validity and acceptability of his or her scientific conclusions.   They can no longer simply claim that they are an “expert” whose opinions and conclusions should be accepted without proof.

An example of a potential Daubert challenge for exposure assessors or industrial hygienists exists in the use of first principle physical-chemical models of human exposure.   A few years ago we were subject to a potential Daubert challenge for some modeling work conducted in anticipation of a court case.  In preparation for this challenge we presented the relevant details of the 2-zone model relative to Daubert in a draft affidavit for the court.   We specifically elucidated the Daubert criteria to address a potential challenge for the Near Field/Far Field (two-zone) indoor air model of breathing zone concentration.  As often happens, the case was settled before we had an opportunity to present our arguments.   We decided to summarize this work in a paper for the benefit of other exposure assessors and Industrial Hygienists for future cases. 

My friends and colleagues Tom Armstrong (fellow modeler) and Michael Taylor (lawyer specializing in occupational health law) published the paper in the Journal of Occupational and Environmental Hygiene as a commentary.   If you want a PDF of this paper I will be happy to send it to anyone requesting it of me at

The paper goes into some considerable detail relative to the two-zone model and the model from which it was derived, the well-mixed box model.   In the paper we present the model theory and discuss our methods, findings, and conclusions below following four key Daubert requirements:

1. Testing of the model
2. Status of peer review and acceptance
3. The technique’s rate of error and standards for the model’s application
4. Personal/professional experience and use of the model.

With regard to “Testing of the Model” we conducted a literature search from which we located and documented multiple studies where the NF/FF technique has been appropriately used and tested. The testing of the model actually involves several stages:

1. Correctness of the mathematical derivation
2. Implementation of the computations and verification of the calculations
3. Testing of the model against suitable real world air concentration data: Verification of model predictions against real world data needs to include situations for the predicted and measured near field concentrations.

Regarding the status of peer review and acceptance, we found and documented multiple peer reviewed publications on the model, coverage in two editions of the AIHA book: Mathematical Models for Estimating Occupational Exposure to Chemicals, and coverage in chapters in two other texts.  In addition, we provided a table that lists numerous studies in a review of the NF/FF model included in a paper by Jennifer Sahmel et al in 2009.  We also included at least 10 others published peer reviewed journal articles using the model to successfully predict exposures to various airborne contaminants.

The models rate of error is actually bound up in its standards for application.   Since all models are idealized portrayals of reality they are accurate to the extent that the reality under investigation conforms with their assumptions.   Indeed, the standards for the model’s application have been stated as the principle “bounding conditions” for the model’s use and can be summarized from the various publications and book chapters as follows:

• The contaminant is instantaneously mixed throughout the near-field and far-field work space.
• There is limited airflow between the two zones.
• The random air velocity between the two zones is uniformly distributed across the NF/FF interface surface.
• There are no significant cross drafts.

Testing the model versus real world concentrations indicates that many scenarios reasonably conform to the above conditions and model predictions typically come within a factor of 0.5 to 2 fold of measured values.   This is good performance for any concentration predicting model.   The paper has all the details and as mentioned above is available to all who ask for it.

No comments:

Post a Comment