U-M Scientists Struggle to Explain Dioxin Study
A stakeholder requested that the U-M Dioxin Exposure Study compare their data to that gathered by Dow and Ann Arbor Technical Services for the Middle Tittabawasse River study. The scientists' reply does not appear to be of high scientific quality.
The problem starts with the data. The U-M study showed 7% of properties with dioxin levels over 1000 ppt, while the Dow study showed 47-67%. In addition, the Dow study had a mean value of 1300 ppt, while the U-M study reported a mean between 237 and 285 ppt.
Quote from the stakeholder request:
The U-M scientists' response is to analyze a subset of their data, and from the subset analysis get numbers within spitting distance of the Dow numbers. This practice is terrible science. They are slicing and dicing their numbers to match somebody else's results, rather than determining what may be flawed about their study. It also raises questions about why they originally reported numbers that were relatively low, and why they were trying to sell the conclusion that dioxin is not a risk to flood plain residents based on their particular choice of how to report the numbers.
Even given that, the reply is not very convincing. They do not report even the number of samples in their subset, nor do they report mean concentrations. Furthermore, it seems unlikely to me that the subset could generate such a high percentage. Either most of the high concentration sites were included in the subset, raising the question of why that would happen, or the subset was not very large, raising questions both about its significance, and about the methodology of the original study in including so few high concentration sites.
The reply is signed by David Garabrant and Alfred Franzblau.
The problem starts with the data. The U-M study showed 7% of properties with dioxin levels over 1000 ppt, while the Dow study showed 47-67%. In addition, the Dow study had a mean value of 1300 ppt, while the U-M study reported a mean between 237 and 285 ppt.
Quote from the stakeholder request:
These discrepancies highlight the agencies' concerns that the UMDES data may not be representative of the population of greatest concern for soil exposure - residents that live along the river in areas that have repeatedly flooded.
The U-M scientists' response is to analyze a subset of their data, and from the subset analysis get numbers within spitting distance of the Dow numbers. This practice is terrible science. They are slicing and dicing their numbers to match somebody else's results, rather than determining what may be flawed about their study. It also raises questions about why they originally reported numbers that were relatively low, and why they were trying to sell the conclusion that dioxin is not a risk to flood plain residents based on their particular choice of how to report the numbers.
Even given that, the reply is not very convincing. They do not report even the number of samples in their subset, nor do they report mean concentrations. Furthermore, it seems unlikely to me that the subset could generate such a high percentage. Either most of the high concentration sites were included in the subset, raising the question of why that would happen, or the subset was not very large, raising questions both about its significance, and about the methodology of the original study in including so few high concentration sites.
The reply is signed by David Garabrant and Alfred Franzblau.
posted by Tommy Times at 7:18 PM
0 Comments:
Post a Comment
<< Home