Multigroup Principal Component Analysis Research Articles

How does drought impact burned area in Mediterranean vegetation communities?

Drought and water stress are widely known to influence fuel moisture content and flammability, although differences do exist according to the response mechanisms and adaptive traits displayed by plant communities. In the Mediterranean basin, as a result of climate change, extreme drought events are expected to become more frequent and severe, envisaging episodes of increased fire risk. In this paper, we expand the scale of analysis on how does drought influence wildfire incidence exploring the joint influence on burned area of drought duration, magnitude and temporal distribution, and the affected vegetation communities (VCs). We leveraged the weekly adaptation of the Standardized Precipitation Evapotranspiration Index (SPEI) and historical fire perimeters to model complex interactions between drought and burned area mediated by VC composition and structure. We applied multivariate factor analysis (multi-group Principal Component Analysis) and non-parametric mixed regression models (GAMM) to a set of 1-to-48 weeks SPEI and 10 dominant VCs. We detected a significant influence of drought events (negative SPEI) on burned area, although differences in terms of seasonal distribution and VC were observed. Drought played a major role in explaining burned area in late spring and autumn by altering the usual positive rainfall-evapotranspiration balance, suggesting a potential lengthening of the fire season given the projected drought trends in the next decades.

Understanding wildfires in mainland Spain. A comprehensive analysis of fire regime features in a climate-human context

Understanding fire regime is a crucial step towards better knowledge of the wildfire phenomenon. However, the concept itself, in spite of its widespread use, still lacks a clear, widely accepted definition and there is no general agreement on which features define it best. In this paper we provide an in-depth characterization and description of fire regimes in three regions – Northwest, Hinterland and Mediterranean – comprising the whole of mainland Spain, to identify their key features. Data on number of fires, burned area, fire season and cause are retrieved from historical fire records for the period 1974–2010. Specifically, fire frequency, burned area, number of natural/human-caused fires, burned area from natural/human-caused fires, number of large fires (≥500 ha), and burned area from large fires were examined for each region and fire season. We used a multi-group Principal Components Analysis approach to determine the importance of each fire regime feature. Next, climate and socioeconomic variables were explored using Multidimensional Scatterplots and Generalized Additive Models to find the extent to which fire regimes are controlled by either environmental, human, or both factors. Results revealed differences among regions and seasons in terms of the characteristics of their respective fire regimes. However, several common features have been identified as key components of fire regimes, regardless of region or fire season: fire frequency, number of large fires, and burned area from natural fires. In addition, results confirm that fire regime in the Northwest area mainly depends on human activity, especially during winter, in contrast to the Mediterranean region.

Assessing the structure of the CAST (Cannabis Abuse Screening Test) in 13 European countries using multigroup analyses.

Our aims are to describe and explain the structure of the Cannabis Abuse Screening Test (CAST) across countries. Standard statistical analyses fail to describe and explain several variables simultaneously while taking account of the group structure of individuals. The 2011 European School Survey Project on Alcohol and other Drugs (ESPAD): 5204 last-year cannabis users aged 15-16 from 13 European countries. Multigroup principal component analysis (mgPCA) and multigroup partial least squares (mgPLS). MgPCA shows that the CAST has a two-dimensional structure (frequency of use/problems and non-recreational use/dependency symptoms). All the countries present a good concordance with the common structure, except Kosovo, Lichtenstein and Romania. MgPLS shows that three explanative variables (in a total of eight) are mainly related with the CAST (the frequencies of cannabis use in the last 12months and in the last 30days and the age at first cannabis use) while Kosovo, Lichtenstein and Romania also present specificities. The CAST structure appears stable in the 13 countries except for Kosovo, Lichtenstein and Romania that also show specific relationships between the CAST variables and their determinants.

Evaluating the novel-environment test for measurement of exploration by bird species

Novel-environment tests are the most widespread experimental technique for characterizing exploration, yet detailed evaluation of their performance among species is lacking. We compared the test for eight bird species by combining three well-known metrics of behavior: movement frequency, proportion of features visited, and scanning. In both overall and species-level analysis of our multi-group principal component analysis, all three metrics loaded strongly and similarly on one principal component, explaining comparable ranges of variation. We conclude that novel-environment tests are a robust means of quantifying exploration and that scanning behavior may be an important but under-used metric for exploration behavior.

Multivariate analysis of multiblock and multigroup data

We address the problem of analyzing one or several blocks of variables measured on the same individuals which are a priori divided into several groups. In this framework, we focus on the within-group analysis. For the case of a single dataset, we consider multigroup Principal Component Analysis proposed by several authors (Levin [18]; Krzanowski [16]; Kiers and Ten Berge [13]). A new optimization criterion which characterizes this method and an extension to the case of multiblock datasets are presented. The method is illustrated on the basis of a dataset pertaining to sensory analysis.

Multi-Site Performance Monitoring in Batch Pharmaceutical Production

Abstract A challenge facing the pharmaceutical and chemical industries is how to understand and identify differences in process behaviour where a product is manufactured at two different sites. Three approaches based on multi-group principal component analysis are investigated and benchmarked against single site models. The multi-group approach is shown to remove differences between sites such as operational scale thereby enabling the analysis to focus on identifying differences in variation between the two sites that are not a consequence of process configurations. From the analysis it is observed that the multi-group approach can assist in the understanding of manufacturing performance.

Viewing Spearman's hypothesis from the perspective of multigroup PCA: a comment on Schönemann's criticism

Jensen's test of Spearman's hypothesis is meant to demonstrate the importance of general intelligence in Black–White (B–W) differences in psychometric intelligence test scores. Schönemann purports to demonstrate, through an analysis of real and simulated data, and the presentation of a theorem, that Spearman correlations are artifacts. We discuss the theorem and conclude that the theorem cannot be advanced in support of the contention that Spearman correlations are positive and substantial by mathematical necessity. The theorem encompasses a multigroup principal component analysis (PCA), which is a viable model to investigate Jensen's proposition that Blacks and Whites differ mainly with respect to g. We view Schönemann's simulation study in the light of the multigroup PCA model, and interpret it as a study of the specificity of Spearman correlations, given model violations. As such, we find it to be open to several criticisms. Schönemann does raise an important issue, namely whether Spearman correlation can be trusted to prove the importance of g in B–W differences in psychometric IQ scores. Based on recent studies, we consider the Spearman correlation to be a suboptimal test of Spearman's hypothesis, and contend that an explicit model-based approach should be used.

Using structural equation modeling to fit models incorporating principal components

The aim of this article is to consider models incorporating principal components from the perspective of structural equation modeling. These models include the principal component analysis of patterned matrices, multiple analysis of variance based on principal components, and multigroup principal component analysis. We demonstrate that these models can be fit readily using the programs LISREL 8 and Mx. The models and certain extensions are discussed, and several illustrations are given.

Principal component analysis using LISREL 8

The object of this article is to demonstrate that the widely disseminated program LISREL 8 can be used to carry out principal component analyses. LISREL 8 offers a number of useful possibilities including multigroup principal component analysis and principal component regression. Several illustrative analyses are reported.

Are Synthyris and Besseya closely related to Wulfenia (Scrophulariaceae – Veroniceae)? Evidence from floral development

Within the tribe Veroniceae (Scrophulariaceae), Synthyris and Besseya have been grouped at times with Wulfenia, because of their rosette habit, and with cauline-leaved Veronica, because of similarities of pollen and seed morphology. Floral development patterns of Synthyris reniformis, Besseya alpina, and Wulfenia carinthiaca were studied with quantitative methods to contribute additional evidence towards resolving the phylogenetic relationships of those genera. For each species, 19 flower organ measurements were taken on each of 50 buds from all stages of development. Compared to the fast early corolla development observed in most Scrophulariaceae, all species examined shared a delay in early corolla tube growth which was less extreme than that observed in Veronica. A strong delay in early corolla lobe growth was shared only between Veronica, Synthyris, and Besseya. This derived attribute is a possible synapomorphy supporting hypotheses of monophyly of those genera. Summarizing all measurements with multigroup principal components analysis showed that the overall floral allometries of Wulfenia have more similarity with those of Digitalis than with Veronica and relatives; this concurs with evidence from vegetative and reproductive morphology. Key words: floral development, allometry, phylogeny.

A Morphometric Analysis of Heterophyllous Leaf Development in Ranunculus flabellaris

Environmentally induced heterophylly provides a useful model system for studying the developmental basis of differing leaf shapes. The semiaquatic Ranunculus flabellaris exhibits typical heterophyllous shape variation with change in water level: aerial leaves have short lobes, while lobes of leaves produced underwater are elongate. Submergence in a 25 μM abscisic acid solution (ABA) produces an aerial-like leaf shape. In this morphometric study, x and y coordinates of nine landmarks of leaf outlines were captured from video images, and linear dimensions between the landmarks were used in multigroup principal components analysis as a method to quantify the timing of developmental divergence in leaf shape of plants grown underwater, in air, or in ABA. Shape differences between aerial and water leaves were detected 16 d after leaf emergence, while size differences were not apparent until after 18 d. This developmental pattern represents a very late structural divergence of leaf forms and additionally indicat...

Bundle sheath and mesophyll cell differentiation in the C4 dicotyledon Atriplex rosea: quantitative ultrastructure

In leaves of most C4 species, both bundle sheath and mesophyll cells are derived from ground meristem, yet at maturity differ in photosynthetic enzyme complement and in cell size, shape, and subcellular ultrastructure. This quantitative ultrastructural study of bundle sheath and mesophyll cell differentiation in Atriplex rosea shows that while developmental pathways of bundle sheath and meosphyll cells are generally coordinated, the timing of developmental divergence differs among individual characteristics. For instance, bundle sheath cells are larger, with more chloroplasts and more and larger mitochondria by 8 days after leaf emergence, while differential growth of mesophyll cell chloroplast peripheral reticulum and increase in thylakoids per granum in bundle sheath chloroplasts do not develop until after 12 days. Multigroup principal components analysis (M-PCA) of the data emphasizes that the greatest source of variation is overall size change as both cell types expand. M-PCA also identifies patterns of allometry within the data; for instance, mesophyll cell vacuoles and chloroplast peripheral reticulum undergo greater relative growth than do bundle sheath microbody area and number. The greater structural specialization of bundle sheath cells is reflected in higher growth rates from the time of divergence, but developmental change in both cell types continues until leaf expansion is complete. Most structural changes occur substantially after the stage of cell-specific expression of C4 enzymes. Key words: bundle sheath, mesophyll, C4 photosynthesis, leaf development, Atriplex rosea, multigroup principal components analysis.

Quantitative Comparison of Floral Development in Veronica Chamaedrys and Veronicastrum Virginicum (Scrophulariaceae)

Development in Veronica and Veronicastrum was studied to elucidate the growth patterns responsible for differences between their mature flowers. Nineteen floral dimensions were measured on buds from initiation to anthesis, and representative stages were recorded with the scanning electron microscope. Bivariate plots indicate the heterochronic changes responsible for the derived flower form of Veronica. The growth patterns of the gynoecium and androecium were little changed. The calyx of Veronica showed slower size increase at the early stages, but continued to grow after that of Veronicastrum had stopped. The most striking change occurred in the corolla lobes of Veronica, where growth was retarded until the beginning of style formation, but afterward was accelerated relative to that of Veronicastrum. The corolla tube of Veronica remained short due to a later onset of growth and slower enlargement later in development. Multigroup principal components analysis (M-PCA) was used to summarize the measurements. The distribution of points along M-PC 1 describes size increase during floral development. Along M-PC 2 the trajectories of the two taxa diverge after the beginning of style growth, while along M-PC 3 they differ from the earliest stages on, corroborating differences observed with the scanning electron microscope. M-PCA can thus be used to portray differences in patterns of floral development, facilitating simultaneous quantitative comparisons of two or more taxa.

The ontogeny of sexual dimorphism in the cranium of Bornean orang-utans (Pongo pygmaeus pygmaeus) as detected by principal-components analysis

The ontogeny of cranial sexual dimorphism in the Bornean orang-utan (Pongo pygmaeus pygmaeus) is examined by means of principal-components analysis (PCA). Normalized first components are called allometry vectors or vectors of relative growth and show that sexual dimorphism is present at all stages of growth. Two patterns of sexual dimorphism are present: (1) sexual differences at age groups 2 and 3 are the result primarily of differences in principal component II scores, reflecting mainly shape-related differences, and (2) age groups 5, 6, and 7 show a trend of stronger size-related shape differences with increasing age in the allometry vector along with decreasing differences in principal component II scores, reflecting an increase in size-related shape differences between the sexes. Age group 4 shows a combination of both patterns. Our results support Shea's hypothesis (1985a) that when using multigroup PCAs in closely related taxa, the allometry vector will generally estimate the shape variation resulting from the extension of common growth allometry patterns (ontogenetic scaling). The second and subsequent components summarize shape variation from slope and intercept differences between the groups, provided that ontogenetic scaling is not solely responsible for all the shape differences present. Subanalyses of those dimensions previously found to show ontogenetic scaling and acceleration follow this pattern well. The total sample provides a pattern whereby ontogenetically scaled dimensions possess a stronger influence over accelerated dimensions but still generally follow Shea's hypothesis. Finally, variously derived coefficients provided several interesting findings: (1) strong evidence was found against multivariate isometry for both the pooled and the separate samples, (2) multivariate allometric coefficients for both sexes follow the general growth pattern of negative scaling in neurocranial dimensions and positive scaling in the viscerocranium, and (3) multivariate slopes have a very high correlation with bivariate slopes relative to the same independent X variable, thereby lending further support to Jolicoeur's (1963a, b) allometry generalization.

Data Set Congruence in Northern Ontario Tamarack (Larix laricina, Pinaceae)

Maternal genotypes and three morphometric data sets (cone morphology, short shoot needle cross-sectional anatomy and shape) are available for samples (N = 8 to 11) from two pop- ulations of tamarack from each of five provenance regions in northern Ontario (North Bay, Kenogami R., Thunder Bay, Fort Frances, Red Lake). Data sets were compared using multivariate methods, including canonical variates analysis and a permutation test for resemblance matrix independence. Data sets were represented as dendrograms based on matrices of distances between populations to test the congruence of data sets with the geographic distances between populations, as well as with each other. The distance functions used were chord distance and Nei's genetic distance calculated from genotype data, Euclidean distances calculated from morphometric data (previously ranged if not commensurate), and Mahalanobis' generalized distances calculated for each data set from pop- ulation centroid multi-group principal components analysis scores. Average-linkage was the single sorting algorithm used for clustering. Dendrograms were compared using five descriptors of their structure that were summarized by means of principal coordinates analysis of the Euclidean distances between dendrograms. The effect of data set on resemblances between dendrograms is considerably greater than that of distance function. Populations from the same region tend to resemble each other less than do populations from different regions, with respect to cone and genetic data; only the needle data were appreciably congruent with the geographic relationships. These results plus earlier ones suggest the probable importance of patterns of dispersal and establishment, inbreeding, and the geographic scale of comparison, as well as selection, for the patterns observed.