Edwin, My issue with such a small sample size would not be low power, but low reliability. If you show me the result for a single replicate, I have no confidence whatsoever that that result is typical for the treatment. Just try calculating the 95% confidence intervals around your results. If you can even find the 95% confidence interval for a treatment with a single replicate, you've done something wrong. I'm surprised only one reviewer complained about your sample sizes and that you didn't need to correct this to publish the paper, especially since it sounds like it would have been logistically simple to increase the replication.
I am also surprised G*power 3 would say that an experiment with just one or two replicates per treatment had near-maximum power. Were you testing for the power to detect differences of the magnitude you actually observed? If so, of course the power would be around 1, if your p-values were low (if your ANOVA actually detected a significant difference, its power to detect that difference must be high). For a power test, you want to know the power of your experimental design to detect the smallest difference you would find biologically meaningful. Jim Crants On Fri, Jul 10, 2009 at 3:49 AM, Edwin Cruz-Rivera < edwin.cruz-riv...@jsums.edu> wrote: > Let me rephrase: out of 12 treatments, 10 had one replicate and 2 had 2. > However, these were not natural lakes or transects in geographic zones > that constrained replication. These were 100 ml bottles on a table. > Sorry for the oversimplification. > > Edwin > > > > Changing a little the topic, I have a question about the statement of > > Edwin. He wrote: > > "If the statistics are grossly inappropriate (for example running an > > ANOVA with 12 treatments, but only 1 or two replicates per treatment), > > adequate peer review was clearly not in place." > > Well, I published a paper in which I used 2 way ANOVA with a total of 18 > > groups and 2 replicates per groups. It was peer reviewed, and one of the > > reviewers complained about my statistics, asking for measurements of > > power, perhaps with the expectation that that particular test would have > > no enough power to draw any conclusions. I used a software to measure the > > power of the test (G*power 3), and found that power was the maximum > > possible (1.00) for the effects due to factors 1 and 2, and 0.99 for the > > interaction effect.Was my test flawed? It was peer reviewed! > > Best, > > > > Matheus C. Carvalho > > > > Postdoctoral Fellow > > Research Center for Environmental Changes > > > > Academia Sinica > > > > Taipei, Taiwan > > > > --- Em qui, 9/7/09, Edwin Cruz-Rivera <edwin.cruz-riv...@jsums.edu> > > escreveu: > > > > De: Edwin Cruz-Rivera <edwin.cruz-riv...@jsums.edu> > > Assunto: Re: [ECOLOG-L] "real" versus "fake" peer-reviewed journals > > Para: ECOLOG-L@LISTSERV.UMD.EDU > > Data: Quinta-feira, 9 de Julho de 2009, 10:37 > > > > I believe one of the original questions was how to discern reputable > > journals from those that publish dubious or biased results...or do not > > accomplish proper peer review. I can point to a couple of red flags that > > can be noticed without too much effort and I have observed: > > > > 1) If the articles in the journal come mostly from the same institution > in > > which the editor in chief is located, chances are the buddy system has > > overwhelmed objectivity...especially if the editor is a co-author in > most. > > > > 2) If orthographic and syntax errors are widespread, probably the review > > process was not thorough. > > > > 3) If the statistics are grossly inappropriate (for example running an > > ANOVA with 12 treatments, but only 1 or two replicates per treatment), > > adequate peer review was clearly not in place. > > > > Now these may look like extreme cases, but I have seen too many examples > > similar to the above to wonder how widespread these cases are. I have > > even received requests to review papers for certain journals in which I > > have been asked to be more lenient than if I was reviewing for a major > > journal. This poses a particular dilemma: Is all science not supposed to > > be measured by the same standards of quality control regardless of > whether > > the journal is institutional, regional, national or international? > > I would like to think it should be... > > > > Edwin > > ------------------------------------------------------------------ > > Dr. Edwin Cruz-Rivera > > Assist. Prof./Director, Marine Sciences Program > > Department of Biology > > Jackson State University > > JSU Box18540 > > Jackson, MS 39217 > > Tel: (601) 979-3461 > > Fax: (601) 979-5853 > > Email: edwin.cruz-riv...@jsums.edu > > > > "It is not the same to hear the devil as it is to see him coming your > way" > > (Puerto Rican proverb) > > > > > > > > > > > > > ____________________________________________________________________________________ > > Veja quais são os assuntos do momento no Yahoo! +Buscados > > http://br.maisbuscados.yahoo.com > > > -- James Crants, PhD Scientist, University of Minnesota Agronomy and Plant Genetics Cell: (734) 474-7478
