[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Ecologic versus case-control
----Original Message-----
From: Field, R. William [mailto:bill-field@UIOWA.EDU]
At 01:29 PM 2/11/2002 -0500, Jim Muckerheide wrote:
Bill,
I've always agreed that case-control is stronger in principle. But the quality of the study governs. Poor radon studies do not become 'good' because they are case-control; they are still bad statistics, as shown by poor statistics and nearly random variations in results. Large eco study is better than poor case-control. Demonstrated by the statistics and the consistency in results over hundreds of individual studies. Science doesn't depend on the semantics.
Jim
--------------------------------
Jim,
I think we found an area of general agreement. I agree, each study should be judged on its own merits. There are well performed ecologic studies and case-control studies. There are also less rigorous ones of each type. However, even a well performed ecologic study serves only to formulate a hypothesis. Hypotheses really need to be tested with either a well designed case-control or prospective cohort study.
<Agreement is ok on the principle, but each study must be judged also on how it reflects the relevant science, both statistics and biology. Cohen has several hundred independent studies, plus many independent studies with the same results. Radon case control studies don't. The Iowa study isn't confirmed, and the poor quality of the data and the poor statistics establishes clearly that it is most likely that a replication of the study would produce a significant departure from the current result. (Could argue that it could go either way, but in fact, such errors tend toward the null, so it is more likely to go in that direction.) In addition they are contradicted by the biology and medicine evidence we do have.
Also, a study with excellent exposure assessment does not need as large a sample size as one with poor exposure assessment. Rosner has shown that the more exposure uncertainty you have, the larger the sample size you need to see an association if one exists.
<Right. Radon case-control studies don't know actual exposures. That substantially limits their so-called case-control advantage in principle. Have no longer removed the most significant variable, but left with a small group with no constraints.>
The degree of accuracy in ecologic data versus a case-control study can hardly be judged against the other design. It is like apples and oranges.
<Here you go back to semantics instead of science. The issue is only in the quality/validity of the statistics. Here it is very important that replications and confirmatory studies address the possibility that a bias could exist in a single study that does not have individual consequences. But arguing that ALL the consistent results of independent studies have the same error is equivalent to arguing that it is possible that materials in macroscopic quantities will likely accelerate upward due to the brownian motion all atoms going "up" at the same time! It's not going to happen. But that doesn't keep Samet et al. from arguing it on behalf of their sponsor interests at EPA, DOE, etc.>
Regards, Jim
If the ecologic data is very accurate and you know there is no cross-level bias or other non linear dependent factors, an ecologic study can be very accurate on the aggregate level.
Bill
************************************************************************
You are currently subscribed to the Radsafe mailing list. To unsubscribe,
send an e-mail to Majordomo@list.vanderbilt.edu Put the text "unsubscribe
radsafe" (no quote marks) in the body of the e-mail, with no subject line. You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/