I've been meaning to comment here too. When I teach statistics, my goal is to give the graduate students a "toolbox" if you will, of useful ways to test ideas. More complex statistics comes later. In teaching, I use the idea of testing hypotheses, with a very important caveat. Both, null and alternative hypotheses have to be biologically sensible and biologically possible. I know I find many published papers that gloss over the null, but it turns out, on deeper inspection, that it was not a possibility and so refuting it was unavoidable.
Apply that idea, that the null also must be reasonable, logical and possible, and you may find that many null hypotheses are none of those. Jim On Thu, Mar 10, 2011 at 13:13, Kevin Mueller <kem...@psu.edu> wrote: > If we iteratively modify our hypotheses through the process of data > collection, data analysis, or manuscript preparation, how different is this > process from "observational" or "exploratory" research? It is, of course, > different to some debatable extent. Regardless, I think Paul's comments shed > light on the reality that there is a large gray area between the extremes of > purely observational studies and purely hypothesis driven studies (which his > 2005 paper apparently documents). Given this, I find the explicit or > underlying claims of superiority made by proponents of hypothesis driven > research to ring false (despite some of the strong benefits of hypothesis > testing that Paul and others have made clear). I find such claims ironic > since the result of many observational or exploratory studies is, gasp, a > hypothesis. > > Finally, regardless of the language we use to reference hypotheses in our > introductions, I ask: Is it always beneficial to cloak studies that are > somewhat exploratory behind a veil of singlular hypothesis testing? Or > might we also sometimes gain and share insights by making the process of > data exploration and hypothesis testing/modification more apparent in our > manuscripts? > > To be clear, my comments are more in response to a general > narrow-mindedness that I've observed among some natural scientists, not to > any particular post or 'poster' in this recent thread (i.e. I found Paul's > post insightful and not especially narrow-minded). > > Kevin Mueller > > On Mar 9, 2011, at 11:00 PM, Paul Grogan wrote: > > "Furthermore, often during the data interpretation or write-up > stage, additional reflection on the processes of experimentation and > evaluation of the data may indicate to the scientist (or to a manuscript > reviewer) that the test did not reflect the hypothesis as well as > originally > thought. In such cases, further refinement or editing of the hypothesis > statement should be made so that the final research output – the > peer-reviewed publication disseminating the new knowledge – is as accurate > and accessible to others as possible. As a result, I usually finish my > manuscript Introduction sections with: “We used our data to test the > following hypotheses....” (rather than “We tested the following > hypotheses... which gives the impression of great foresight on the part of > the author)."
