|Cason Jr, John|
Submitted to: Poultry Science Association Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 5/16/2006
Publication Date: 7/16/2006
Citation: Cason Jr, J.A., Pesti, G.M. 2006. Mean separation procedures used in poultry science [abstract]. Poultry Science Association Meeting Abstract. 85(1):125. Interpretive Summary:
Technical Abstract: Research papers published in Poultry Science were studied to determine the frequency and appropriateness of use of procedures to test for significant differences between treatment means. Excluding papers from symposia, 255 papers and research notes were published in Poultry Science in 2004, of which 198 or 77.6% reported differences between means of 3 or more treatments. Mean separation was accomplished with the following procedures: Duncan’s multiple range test, 46 papers; least significant difference, 32; Tukey’s, 25; t test, 24; contrasts, 21; least squares, 21; paired t test, 12; and other, 20. Eleven papers showed separation of means in tables or figures but did not specify a method. Use of more than one procedure in a single paper resulted in a total of 201 procedures used in 187 papers that specified methods. Some methods were not used appropriately. Examples of inappropriate use of Duncan’s multiple range test, for example, included use with factorial designs (8 instances), application to designs with graded levels of treatments where regression techniques could have been used (3), use in experiments designed with internal structure such that treatments could have been analyzed with orthogonal contrasts (6), and in cases where the only comparisons discussed were between individual treatments and a control (4). Despite numerous papers and textbooks that discuss the proper use of mean separation procedures, many papers in Poultry Science report results that were not analyzed by methods most likely to find significant and meaningful differences in the data under study.