Friday, November 16, 2007

Brenna's Testimony - Questions

Dr Brenna is an expert in the field of IRMS and has had many papers published in this area. Some have been referenced in Dr Meier-Augenstein's own papers. However, some unusual anomalies appeared during the course of the hearing. Anomalies in the form of differences of opinion between Brenna and Meier-Augenstein over the answer to relatively straightforward questions (the effect on o/oo values for overlapping peaks).

[MORE]


This naturally leads one down the path of looking for other anomalies. After all, their difference of opinion over the effect of overlapping peaks is one which forces an observer to believe either one person or the other. They cannot both be correct. Their opinions represent the compliment of each other.

Brenna testified that during IRMS processing of the B samples, the process was automatic and that the only intervention by the technicians was for what he referred to as "quality control" purposes. Clearly, the question had been asked to determine whether the results could have been influenced by the technicians (no doubt in light of the fact that they knew which were Floyd's samples and which were the blanks). Brenna testified that the process had been automatic, not manual. Thus implying that the they could not have influenced the results. That effectively removed any question of bias during this process and, in the eyes of the arbitration panel, validated all of those results.

Moving to page 169 of the official transcript, Brenna describes the quality control actions carried out by the technicians, as observed by himself during the official testing of all of Landis' B samples. His testimony describes how the LNDD technicians, following their internal SOP, determine for themselves where the peak of interest is deemed to start and end. This is done by observing the 45/44 ratio plot and the value of the background o/oo at the start of the peak (more on this later). The first observation one would make at this point is that, clearly, selection of the integration limits are not done automatically. So which aspect of this process was automatic ?. Defining where the peak integration begins and ends also defines the background removal (everything below a straight line fit between the intersection of the integration limits and the peak). So in defining the integration limits, they are defining what is to be considered background. So which aspect of this process was automatic ?. All of the decisions which influence the computed o/oo value of the peak of interest have been taken by the lab technician. The only thing they didn't do was the number crunching necessary to evaluate their decisions. We'll leave this as an open question. Was this an automatic process ?. Brenna testified that it was.

The SOP for determining the peak start and stop times is described in detail in Brenna's testimony. The o/oo value of the background is determined from the 45/44 ratio plot at the start of the peak. The 45/44 plot is then scanned to determine where the same 45/44 value reoccurs at what appears to be the end of the peak. This is then deemed to be the end of the integration interval. A sanity check is performed by checking the defined interval on th 44 plot and any gross errors will be observed there. However, the "integration for Idiots" series highlighted that gross errors are not required to skew the results.

The presence of interference between peaks distorts the true 45/44 ratio plot for a peak. This can lead to shifts in when the 45/44 plot appears to return to the baseline. So, by following the LNDD SOP (in the presence of the sort of interference between peaks evident in Floyd's F3 plots), inaccurate results become more than just theorhetical possibilities. They become hard to avoid realities.

The most obvious question was how unbiased was Brenna's testimony ?

0 comments: