Daubert in Engineering Including Biomedical Engineering
I recently cross examined an expert witness, a biomedical engineering professor, who had testified that my client could not have fallen in the way she said she did. He had physics based free body diagrams, citations to peer reviewed published papers (including several that he had written) that he cited as supporting his expert testimony. The one of these papers that he relied most comprehensively upon even had a multiple regression type of statistical analysis like is discussed on this site. He used a MANCOVA model, a type of multiple regression analysis, to estimate -fall direction and impact velocities of an accident victim. He was a distinguished one-time Ivy League professor, widely published.
He used a physics-based model to explain that my client could not possibly have fallen the way that she said she fell. He had published 100 peer-reviewed papers, had physics based free body diagrams and was sure he was in control of the room.
About five questions into my cross I asked him how he had tested that the residuals of his regression model were normally distributed. (A key part of the standards for the use of regression analysis).
HUH? He said.
How do you test that the residuals of your regression model were normally distributed?
Well, I HUH? He said
How did you test the residuals?
Bluster here Bluster there, . . . Well it was a peer-reviewed paper he said.
But how did you test the residuals?
Well this went on a couple more rounds,but ultimately he admitted that he could not demonstrate how he checked that the residuals of his model were normally distributed, nor could he explain other details of his statistical method. He could not to establish that the normality assumption was satisfied, and this invalidated his tests and error rates, rendered his model unreliable according to scientific inquiry and (at least) Daubert’s factor 5. There were plenty of other things going on in the litigation, but the case settled immediately after the deposition. I had flown to Oregon on to take this expert’s deposition, Alaska Airlines lost my luggage in San Francisco, and this professors client settled with mine before Alaska Airlines found my suitcase.
The point of this case is do not sleep on Daubert’s factor 5. Just because there are statistical tests behind en expert’s testimony does not mean that they are unbiased or proper or done in accord with the standards for their use that Daubert mandates.
All the testing, error rate analysis and all the rest goes out the window if the expert does not follow the standards for the use of the technique.
The second point is that many Plaintiff’s lawyers think Daubert stacks the deck against them. Well, on the day of that cross, I was a plaintiff’s lawyer and I dinged (crushed) a defense expert.
I work with plaintiffs and Daubert doesn’t stack the deck against plaintiff’s lawyers so much as it stacks the deck against junk testimony and lawyers that don’t command Daubert.