Required Reading

Introduction to error and good research design

Pease, C.M. and J.J. Bull. 2000. Science for Business, Law and Journalism. Read Chapter 9 Error is unavoidable.

Examples of error in two scientific papers

Grandjean, P. et al. 1998. Cognitive performance of children prenatally exposed to "safe" levels of methylmercury. Environmental Research, Section A 77: 165-177. Note: You have already seen this paper. Once again, do not read it from front to back. Rather, read it with a view towards answering the questions posed below. Whereas we discussed models and impediments previously, in this class our focus will be on different types of error.

Schober, S.E. et al. 2003. Blood mercury levels in US children and women of childbearing age, 1999-2000. JAMA 289: 1667-1674.

Questions for class discussion

Note that we will spend much of the class on question 1. below. I also have a few simple class demonstrations to help illustrate error in scientific data.

Error

1. Drawing on Grandjean et al. (1998) and Schober et al. (2003), give examples of the 4 types of error in Chapter 9 of Science for Business, Law and Journalism.

2. All data is, at some level, in error. So scientists must necessarily used erroneous data. But how do scientists determine if the errors are large enough to matter? In other words, when can we ignore these errors, and when must we pay attention? And how do we decide? Hint: We characterize the size of the error: Is it big or is it small?

3. Find an example from Grandjean et al. (1998) or Schober et al. (2003) that demonstrates that "data are models." For your example, state what your data (i.e. model) substitutes for. What assumptions does your data (i.e. model) make? Is there any data that is not a model? Restated, can a scientist ever gather data, and know with absolute certainty that the data are fully correct?

Schober et al. (2003)

4. What are the goals, study design and conclusions of Schober et al. (2003)?

5. What hypothesis do Schober et al. test? Do they state this hypothesis explicitly? Do the authors attempt to falsify a hypothesis?

6. We previously discussed a four part risk assessment paradigm as outlined in the Red Book. How does the Schober et al. paper fit into this framework?