Fræðaþing landbúnaðarins - feb. 2010, Síða 302

| FRæðAÞING LANdBúNAðARINS 7, 2010302
tank roughly doubles the statistical power while an increase from n of 50 to 100
increases power by 918. Further increases in n have a limited effect on effect size.
This may suggest that the optimum design of growth studies is to test the treatments
in triplicates or quadruplicates with 50100 fish in each tank.
The Cerror ranged between 1556 and Cwithin was 012. These are similar
values as reported by Ling & Cotter (2003) for growth studies on Atlantic salmon.
The variance in data on Atlantic cod and tilapia were larger than in data on turbot and
Arctic charr. It is not clear if this difference reflects random occurrences between
studies on the same species or if this reflects actual differences among species
possibly caused by differences in behaviour and individual competition or human
error. Further investigations are suggested to ascertain this.
In conclusion, we suggest that treatments in growth studies should be tested in at least
triplicates to minimize effect size while maintaining adequate statistical power. There
is little gained in having a sample size of more than 100 individuals. Although some
differences may exist in the variance of data on body mass in different species of fish,
the general conclusions of this study may apply to other species.

The authors wish to thank the research team at Holar University College and Messrs
Maurice Ssebisubi and Lazaro W. Jere of Bunda College, University of Malawi; for
providing their experimental data. Godfrey Kubiriza was funded by the United
Nations University, Fisheries Training Program.

António T. Álvaro R. & Teresa C. 2009. Statistical Power Analysis with Microsoft Excel: Normal
Tests for One or Two Means as a Prelude to Using NonCentral Distributions to Calculate Power.
(1), 1 – 21.
Araujo, P. & Frøyland, L. 2007. Statistical power and analytical quantification.  
 , 305–308.
Deng, H. 2005. Does It Matter If NonPowerful Significance Tests Are Used in Dissertation Research?
  (16), 1 13.
Festing, M. F. W. 2006. Design and Statistical Methods in Studies Using Animal Models of
Development. ,  (1), 514.
Hoenig, J. & Heisey, D. M. 2001. The Abuse of Power: The Pervasive Fallacy of Power Calculations
for Data Analysis.  , 16.
Imsland, A.K., Gústavsson, A., Gunnarsson, S., Foss, A., Árnason, J., Jónsson, A., Smáradóttir, H.,
Arnarson, I. & Thorarensen, H. 2008. Effects of reduced salinities on growth, feed conversion
efficiency and blood physiology of juvenile Atlantic halibut (  L.).
, 254259.
Imsland, A.K., Gunnarsson, S., Ásgeirsson, Á., Kristjánsson, B., Árnason, J., Jónsson, A.F.,
Smáradóttir, H. & Thorarensen, H. 2010. Long term rearing of Atlantic halibut at intermediate
salinities: effect on growth and blood physiology. , 19.
Ling, N. E. & Cotter, D. 2003. Statistical power in comparative aquaculture studies. 
 , 159–168.
Ling. N. E. 2007. Efficient analysis of growth trial data.  , 728 – 732.
Rosenthal, R. 1990. Replication in Behavioral Research. In Neuliep, J.W. (Ed.).  
 [Special Issue.] 
 5 (4), 130.
Thomas L. 1997. Retrospective Power Analysis. Conservation Biology 11(1), 276280.
Scheffé, H. 1959. The analysis of variance. John Wiley & Sons, New York, USA.
Whalberg, H. J. 1984. Improving the productivity of America's schools. Educational Leadership 41(8),
1927.