On Canonical Variates Analysis and the Construction of Confidence Ellipses in Systematic Studies

Owen, James G.; Chmielewski, Margaret A.

doi:10.1093/sysbio/34.3.366

Cited by 33 publications

(24 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the BEAST approach, several steps are typically used to assign reference-and test-site data into appropriate classes . The last step in the model is construction of a Gaussian bivariate probability ellipse (Altman 1978, Owen andChmielewski 1985) around reference sites in nonmetric or hybrid multidimensional scaling ordination space to assess whether assemblage composition at a test site lies outside of what would be expected within the reference condition. However, we only had 1 reference stratum and a substantial amount of published information from our study area to help assign plots as either reference or impaired (transition and impacted) a priori.…”

Section: Discussionmentioning

confidence: 99%

Evaluating Subsampling Approaches and Macroinvertebrate Taxonomic Resolution for Wetland Bioassessment

King¹,

Richardson²

2002

Journal of the North American Benthological Society

152

114

View full text Add to dashboard Cite

Abstract. Methods for wetland bioassessment using macroinvertebrates are not well developed. Two of the most controversial issues in stream bioassessment, subsampling and taxonomic resolution, have yet to be quantitatively addressed for wetlands. Using a multivariate approach, we evaluated the efficacy of family-, genus-, and species-level assemblage data in reflecting the environment and distinguishing impaired sites from the reference condition. We used 5 basic levels of subsampling (100-, 200-, and 300-organism fixed counts; 10% and 25% fixed areas), an integrated subsample requiring a minimum fixed count and fixed area (100&10%), and 100-count and 10%-area subsamples coupled with a supplementary large-rare (LR) search. Data were obtained from 1.5-m 2 composite samples collected from 126 plots along a 10-km-long eutrophication gradient in the Florida Everglades. Our results suggest that effectiveness of subsampling depended more upon the minimum number of individuals retained than minimum area or proportion of the sample picked. Fixed-area subsamples were generally less efficient than fixed counts, with 200-and 300-individual fixed counts resulting in significantly greater assemblage-environment relationships and much higher accuracy in detecting impairment than 10% fixed area, despite averaging similar numbers of individuals. The greatest improvement with increasing subsample size was observed between fixed counts of 100 and 200 individuals; detecting impairment, in particular, was not markedly improved with subsample sizes Ͼ200 individuals. Supplementing subsamples with a LR search resulted in only very slight improvements in assemblage-environment relationships, but was effective in improving prediction accuracy, particularly for family-level data. However, family-level assemblage-environment relationships and abilities to detect impairment were inferior to genus-and species-level data, regardless of subsample size. Species-level data performed best, primarily because of the large proportion (Ͼ20%) of total species belonging to Chironomidae. The potential importance of Chironomidae to wetland bioassessment was further revealed through an evaluation of a tiered-taxonomic approach, which showed that non-Chironomidae family-level data tiered with species-level Chironomidae data produced results very similar to those obtained using genus-or species-level data exclusively. Our results suggest that fixed counts Ն200 or integrated fixed-area/fixed-count approaches that consistently obtain a minimum of 200 individuals should be considered as minimum subsample sizes for wetlands. We additionally advocate LR searches and recommend genus-or species-level taxonomy, particularly for the Chironomidae.

show abstract

Section: Discussionmentioning

confidence: 99%

Evaluating Subsampling Approaches and Macroinvertebrate Taxonomic Resolution for Wetland Bioassessment

King¹,

Richardson²

2002

Journal of the North American Benthological Society

152

114

View full text Add to dashboard Cite

show abstract

“…2. Statistical bivariate variability of quadrat populations about their centroids is depicted by 95% confidence ellipses produced as described by Owen and Chmielewski (1985). productivity hypotheses do not hold up as general explanations ofbroad geographic patterns of mammalian species richness.…”

Section: Species Richness Patternsmentioning

confidence: 98%

An Analysis of the Spatial Structure of Mammalian Distribution Patterns in Texas

Owen¹

1990

Ecology

View full text Add to dashboard Cite

I examined mammalian distribution patterns and species richness in the state of Texas, USA, along two major environmental gradients. I drew a system of square quadrats on a map of the state. I recorded the presence of absence of each native mammalian species and fifteen environmental variables. I analyzed these data using classification and ordination procedures and by statistical regression methods. The purpose of the analysis was to determine if a hierarchical classification explains the spatial structure mammalian distributions well, or if they are best understood as a continuum resulting from the individual responses of species to history and environmental conditions. I used quadratic response surface analysis to relate changes in species richness to environmental gradients and to test ecological hypotheses that richness is explained by climatic stability and primary productivity. Correlation analysis of community ordination axes with a suite of environmental variables indicated that mammalian species distributions differentiate along two major clines: a dominant east—to—west gradient of decreasing precipitation and productivity, and a south—to—north gradient of decreasing mean annual temperature, increasing winter cold, and increasing seasonal variation of temperature. Mammalian community—units, as identified by hierarchical cluster analysis, are essentially continuous with each other. This interpretation is reinforced by the substantial boundary overlap of 95% confidence ellipses about the association centroids. Quadrat ordination scores, contoured on a map of the study area, exhibited rather even patterns of species turnover in both east—to—west and north—to—south directions. This indicates that there are no zones of rapid species replacement followed by areas of faunal homogeneity. The distributions of species that range widely within quadrat sets show little fidelity to the areas to which they are most highly associated. Species appear to be geographically distributed in an individualistic manner according to the vicissitudes of history and their own ecological tolerances. These findings are consistent with the continuum concept community structure for mammals. Quadratic response surface analysis of species richness (bats, rodents, carnivores, and all mammals) with environmental variables estimating productivity and stability of temperature did not support either the productivity or stability hypotheses of species richness. The relationship of these four taxa to a resource gradient of increasing productivity shows that the curvilinear models used by Grime (1979), Huston (1979), and Tilman (1982) for plants apply also to mammals.

show abstract

“…We used principal components analysis (PCA) [23] to find a 95% [13] confidence ellipse surrounding the spread of launch landing points produced in the launch phase. PCA allows us to determine the two-dimensional vector that contains the greatest variance in the distribution of launch points.…”

Section: Launch Error Analysesmentioning

confidence: 99%

Characterizing user performance with assisted direct off-screen pointing

Ens

Ahlström

Cockburn

et al. 2011

Proceedings of the 13th International Conference on Human Computer Interaction With Mobile Devices and Services

View full text Add to dashboard Cite

The limited viewport size of mobile devices requires that users continuously acquire information that lies beyond the edge of the screen. Recent hardware solutions are capable of continually tracking a user"s finger around the device. This has created new opportunities for interactive solutions, such as direct off-screen pointing: the ability to directly point at objects that are outside the viewport. We empirically characterize user performance with direct offscreen pointing when assisted by target cues. We predict time and accuracy outcomes for direct off-screen pointing with existing and derived models. We validate the models with good results (R 2 ≥ 0.9) and reveal that direct off-screen pointing takes up to four times longer than pointing at visible targets, depending on the desired accuracy tradeoff. Pointing accuracy degrades logarithmically with target distance. We discuss design implications in the context of several real-world applications.

show abstract

On Canonical Variates Analysis and the Construction of Confidence Ellipses in Systematic Studies

Cited by 33 publications

References 0 publications

Evaluating Subsampling Approaches and Macroinvertebrate Taxonomic Resolution for Wetland Bioassessment

Evaluating Subsampling Approaches and Macroinvertebrate Taxonomic Resolution for Wetland Bioassessment

An Analysis of the Spatial Structure of Mammalian Distribution Patterns in Texas

Characterizing user performance with assisted direct off-screen pointing

Contact Info

Product

Resources

About