Error in geometric morphometric data collection

Combining data from multiple sources

Bibliographic Data

ID8315871
AuthorsChris Robinson (0000-0001-8896-3538, Department of Biological Sciences Bronx Community College (CUNY) Bronx New York 10453, corresponding author), Claire E Terhune (0000-0002-5381-2497, Department of Anthropology University of Arkansas Fayetteville Arkansas)
Year2017
Volume164
Issue1
Pages62-75
Publication date2017-09-01
Peer ReviewedYes
Open AccessYes
TypeARTICLE
VenueAmerican Journal of Physical Anthropology (JOURNAL)
Journal identifiersISSN: 0002-9483 • E-ISSN: 1096-8644
PublisherWiley (PUBLISHER • GB)
DOI10.1002/ajpa.23257
PMID28573732
OpenAlexW2621195210
LanguageEN
Citations received19
References cited33

OBJECTIVES: This study compares two- and three-dimensional morphometric data to determine the extent to which intra- and interobserver and intermethod error influence the outcomes of statistical analyses. MATERIALS AND METHODS: Data were collected five times for each method and observer on 14 anthropoid crania using calipers, a MicroScribe, and 3D models created from NextEngine and microCT scans. ANOVA models were used to examine variance in the linear data at the level of genus, species, specimen, observer, method, and trial. Three-dimensional data were analyzed using geometric morphometric methods; principal components analysis was employed to examine how trials of all specimens were distributed in morphospace and Procrustes distances among trials were calculated and used to generate UPGMA trees to explore whether all trials of the same individual grouped together regardless of observer or method. RESULTS: Most variance in the linear data was at the genus level, with greater variance at the observer than method levels. In the 3D data, interobserver and intermethod error were similar to intraspecific distances among Callicebus cupreus individuals, with interobserver error being higher than intermethod error. Generally, taxa separate well in morphospace, with different trials of the same specimen typically grouping together. However, trials of individuals in the same species overlapped substantially with one another. CONCLUSION: Researchers should be cautious when compiling data from multiple methods and/or observers, especially if analyses are focused on intraspecific variation or closely related species, as in these cases, patterns among individuals may be obscured by interobserver and intermethod error. Conducting interobserver and intermethod reliability assessments prior to the collection of data is recommended

Biology · Estimator · Principal component analysis · Statistics · Amphibian and Reptile Biology · Computer Science · Evolution and Paleontology Studies · Mathematics · Morphological variations and asymmetry

Tools

Open DOI