MEASUREMENT ERROR ASSOCIATED WITH EXTERNAL MEASUREMENTS COMMONLY USED IN SMALL-MAMMAL STUDIES

• Published in: Journal of mammalogy Vol. 87; no. 2; pp. 216 - 223
• Main Authors: Blackwell, Grant L., Bassett, Suzanne M., Dickman, Chris R.
• Format: Journal Article
• Language: English
• Published: Provo, UT American Society of Mammalogists 01-04-2006; Brigham Young University, Department of Zoology; Oxford University Press
• Subjects: Accuracy; Analysis of variance; Biological and medical sciences; Biological taxonomies; dasyurid; Error rates; Fundamental and applied biological sciences. Psychology; Gender discrimination; mammal; Mammalia; Mammalogy; Mammals; marsupial; Marsupials; measurement error; Measurement errors; model II analysis of variance; morphometrics; observer bias; Population ecology; precision; rodent; Rodents; Specimens; Statistical variance; Variance analysis; Vertebrates: general zoology, morphology, phylogeny, systematics, cytogenetics, geographical distribution; Zoology; Methodology; precision; accuracy; morphometrics; Micromammals; Vertebrata; Mammalia; dasyurid; mammal; model II analysis of variance; rodent; External; observer bias; marsupial; Measurement method; Morphometry; Measurement error
• ISSN: 0022-2372; 1545-1542
• DOI: 10.1644/05-MAMM-A-215R1.1

Measurement error can occur at both the intra- and interobserver level, result in significant biases in collected data, and increase the likelihood of type II errors. We estimated the intra- and interobserver measurement error associated with 4 commonly used and 1 modified external metric in 4 representative Australian small-mammal species, using both preserved laboratory specimens and live wild-caught individuals. Measurement error was larger in live individuals and was significant in a number of cases in metrics that were difficult to define, were highly flexible, or were relatively small. Observer experience significantly affected precision, and overall, our modified hind foot metric (length of the hind foot, excluding the toes) performed similarly to the currently used metric. The number of replicate measurements significantly affected the size of detectable differences between observers, and we recommend that at least 2 replicate measurements per metric be taken as a matter of course.