24.07.2013 Views

February 15-18, 2009 Washington State Convention Center Seattle ...

February 15-18, 2009 Washington State Convention Center Seattle ...

February 15-18, 2009 Washington State Convention Center Seattle ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

MULTIPLE COMPARISONS OF MEANS VS. REGRESSION ANALYSIS IN AQUACULTURE<br />

RESEARCH<br />

Steeve Pomerleau<br />

Aquaculture/Fisheries <strong>Center</strong><br />

University of Arkansas at Pine Bluff<br />

Mail Slot 4912<br />

1200 North University Drive<br />

Pine Bluff, AR 71601 USA<br />

spomerleau@uaex.edu<br />

Comparing treatment means by testing for statistical significant differences is probably the most common approach used in<br />

aquaculture research. However, in several common types of aquaculture experiments such as in fertilization experiments or<br />

stocking density experiments, where treatments correspond to several levels of a quantitative or continuous variable, comparing<br />

treatment means for significant differences is not an appropriate and efficient way of analyzing the data and of obtaining<br />

meaningful information. In those types of experiments, regression analysis should be used to estimate and identify trends or<br />

relationship instead of simply comparing means to find which treatment gave the best results. Unfortunately, in the literature,<br />

multiple comparisons of treatment means are too often unwarranted and the regression analysis is too often ignored where it<br />

would have been the most appropriate approach for the factors being studied. Comparing treatment means is not necessarily<br />

wrong; it is just that more valuable information could sometime be attained with the same amount of resources if a regression<br />

analysis and a slightly different experimental design were used.<br />

This problem has been raised several times in the literature. Chew (1976) wrote: “Duncan’s multiple range test (MRT) is very<br />

often inappropriately used to compare treatments that are factorial in nature or that correspond to several levels of a quantitative<br />

or continuous variable.” Lowry (1992) wrote: “Pairwise, multiple comparisons are appropriate only for comparing unstructured,<br />

qualitative treatments.” Dawkins (1983) wrote: “How do we persuade biologists of the futility of significance testing<br />

between mean responses (ordinate value of y), at series of dose-rates (abscissa values of x), along response curves?” Knud-<br />

Hansen (1997) wrote: “Although the statistical community is quite clear on this point, the aquaculture community (among<br />

other scientific disciplines) has yet to appreciate the inappropriateness and gross inefficiency of multiple range test usage in<br />

analyzing structured experiments.”<br />

2

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!