Hierarchical Multiple Factor Analysis Research Articles

The role of previously glaciated landscapes in spatiotemporal variability of streamflow in snow-dominated watersheds: British Columbia, Canada

Study regionTwenty-six glaciated basins on the Interior Plateau (56° N, 125° W) in British Columbia, Canada. Study focusLandscape and streamflow characteristics in glaciated basins show high spatiotemporal variability, while the linkages among runoff characteristics, landscapes, and glacial history remain unclear. We compiled a 10-year dataset of daily climate and streamflow observations, and 54 landscape attributes for 26 glaciated basins to explore the relationship between landscape, streamflow and glacial history. A Hierarchical Multiple Factor Analysis was performed to analyze landscapes. The hydrological signatures were selected to describe streamflow characteristics. A correlation analysis and machine learning models were used to explore the relationships between the glacial legacy and streamflow variation. New hydrological insights for the region(1)The landscape is well-structured. The principal components of PC1 and PC2 explain 44% of the total landscape variance, PC1 representing basins with steep, high-relief terrain and alpine glaciation and PC2 representing large, elongated basins with moraine and glacial lakes; (2) The landscape heterogeneity has a key role in accurate streamflow predictions. Most hydrological signatures are controlled by landscape components PC1 and PC2 and can be well predicted by Random Forest using landscape features as inputs. (3) Alpine glaciation results in rapid runoff generation and variable flow regime in mountains, while valley glaciation leads to smooth hydrographs with large baseflows and mild and infrequent extreme flow events in plateaus.

From Fermented Wash to New Make Spirit: Assessing the Evolution of Flavour Characteristics of Scotch Whisky Using Lab-Scale Process Simulations

New product development for distilled spirits frequently involves experimental trials at the laboratory scale that attempt to replicate industrial-scale production processes. This process is time-consuming and limits the number of samples that can be analysed. The aim of the present study was to conduct laboratory-scale Scotch malt whisky production experiments to determine if samples taken from earlier in the production process, that is, directly after fermentation (wash stage) or after a single distillation (low wines stage), showed similar analytical differentiation compared to samples of fresh distillates (new make spirits). Napping, a rapid sensory method, was used to assess the impact on flavour characteristics while solid-phase microextraction (SPME) gas chromatography–mass spectrometry (GC–MS) was used to analyse volatile flavour congeners. Hierarchical Multiple Factor Analysis was used to compare the product maps from samples taken at each whisky production stage and revealed a pattern of differences across the samples that could be tracked through the process. Although the flavour descriptors and volatile congeners composition changed at each stage, there were only marginal changes in the differentiation between samples, resulting in the same sample groups being found in all analyses. RV coefficients >0.90 for all analytical comparisons and >0.74 overall showed that all product maps were highly similar to each other and showed the same overall differentiation between samples. These results indicate that the analysis of fermented malt whisky wash may provide sufficient information to proceed to larger-scale industry trials, saving time and allowing a greater number of parameters to be explored.

Instrumental classification of beer based on mouthfeel

Beer is one of the most consumed alcoholic beverages in the world. Classification of beer helps the consumer to find a preferred beer. Sensory assessments of taste are commonly done by sensory panels and therefore susceptible to subjectivity. Mouthfeel is an important parameter for the total perception of beer flavor. Three dimensions of mouthfeel are distinguished: contracting, coating, and drying. In this study 24 beer samples were evaluated chemically. The data were matched with sensorial data obtained from a trained panel. Different chemical analyses were performed; total acidity (TA), total flavonoids (TF), total polyphenols (TPC), total sugars (TS), color, pH, carbon dioxide content, ethanol, bitterness units (BU) and total iso-α-acids (TIA). The data were analyzed by performing several statistical techniques such as analysis of variance, principal component analysis, agglomerative hierarchical cluster analysis and multiple factor analysis. Sensory data obtained from trained panelists on the different mouthfeel attributes correlated with the data found instrumentally. The drying dimension could be expressed using the TPC, BU, TIA and pH. Contracting compounds correlated positively with TA and negatively with pH. As expected, ethanol was strongly associated with burning sensations and carbon dioxide with carbonation. The results of this experiment indicate that commercial beers can be classified into three mouthfeel attributes: drying, coating and contracting. The Principal Component Analysis (PCA) in this study confirmed the dimensions of the mouthfeel model and identified drying and coating as opposites, contracting forces interact on these dimensions. Moreover, these attributes were shown to be quantifiable by instrumental analysis which suggests that a data-driven approach based on mouthfeel could reduce subjectivity in the analysis of taste.

Development of a protein supplemented fruit smoothie using pea protein isolate as a plant-based protein alternative

Protein products are currently among the most important trend products. The product range is becoming more and more versatile. Consumers are also increasingly demanding proteins derived from plant-based sources. In addition to protein products, smoothies are also very popular, as smoothies are proving to be a quick and easy alternative for people on the go. The consumption and sales of smoothies have been growing continuously for several years and new products are constantly being added. The present work proposes a basis for the development of a vegan fruit smoothie whose protein content can be advertised with the claim “source of protein” or “high protein content”. For this purpose, fruit smoothies were enriched with different concentrations of pea protein isolate. Using the check-all-that-apply method, the perception of the sensory properties was evaluated by 67 untrained test persons. The consumers also hedonically evaluated the products using a 7-point scale. In terms of sensory properties, the enrichment of the fruit smoothie with pea protein, depending on the added protein concentration, results in the smoothies being described with attributes such as furry, mealy/coarse and blunt. Correlating the hedonic and analytical data, it was found that these attributes have a negative impact on consumer acceptance. Penalty analysis was used to determine drivers of liking and to identify guidance and possibilities to optimize the product in future work. However, using agglomerative hierarchical clustering and multiple factor analysis two consumer groups have been identified who liked the protein-enriched smoothies. The smoothie with the highest protein concentration was even preferred by one of the two consumer groups, indicating that a vegan smoothie enriched with pea protein has potential among consumers.

Using hierarchical unsupervised learning to integrate and reduce multi-level and multi-paraspinal muscle MRI data in relation to low back pain

PurposeThe paraspinal muscles (PSM) are a key feature potentially related to low back pain (LBP), and their structure and composition can be quantified using MRI. Most commonly, quantifying PSM measures across individual muscles and individual spinal levels renders numerous separate metrics that are analyzed in isolation. However, comprehensive multivariate approaches would be more appropriate for analyzing the PSM within an individual. To establish and test these methods, we hypothesized that multivariate summaries of PSM MRI measures would associate with the presence of LBP symptoms (i.e., pain intensity).MethodsWe applied hierarchical multiple factor analysis (hMFA), an unsupervised integrative method, to clinical PSM MRI data from unique cohort datasets including a longitudinal cohort of astronauts with pre- and post-spaceflight data and a cohort of chronic LBP subjects and asymptomatic controls. Three specific use cases were investigated: (1) predicting longitudinal changes in pain using combinations of baseline PSM measures; (2) integrating baseline and post-spaceflight MRI to assess longitudinal change in PSM and how it relates to pain; and (3) integrating PSM quality and adjacent spinal pathology between LBP patients and controls.ResultsOverall, we found distinct complex relationships with pain intensity between particular muscles and spinal levels. Subjects with high asymmetry between left and right lean muscle composition and differences between spinal segments PSM quality and structure are more likely to increase in pain reported outcome after prolonged time in microgravity. Moreover, changes in PSM quality and structure between pre and post-spaceflight relate to increase in pain after prolonged microgravity. Finally, we show how unsupervised hMFA recapitulates previous research on the association of CEP damage and LBP diagnostic.ConclusionOur analysis considers the spine as a multi-segmental unit as opposed to a series of discrete and isolated spine segments. Integrative and multivariate approaches can be used to distill large and complex imaging datasets thereby improving the clinical utility of MRI-based biomarkers, and providing metrics for further analytical goals, including phenotyping.

Why Not Travel to Malaysia? Perspectives of Future Senior Tourists toward Senior-Friendly Destination Attributes

ABSTRACT The push-pull motivation theory was applied to examine the role of travel motivation in the relationship between senior-friendly destination attributes and future seniors’ travel intention to a senior-friendly destination. A quantitative cross-sectional study was conducted on a sample of 209 future senior tourists in Malaysia. Hierarchical multiple regression and factor analysis results revealed four important attribute dimensions of a senior-friendly destination from the perspective of future senior tourists, i.e., ‘senior-friendly facilities and accessibility,’ ‘tourism resources and attractions,’ ‘complementary services,’ and ‘safety and security services.’ ‘Tourism resources and attractions’ was found to be the only destination attribute that significantly predicts future seniors’ travel intention. Travel motivation was tested to determine its effect on travel intention, however the findings indicated that future senior tourists’ travel motivation has no moderating role in the relationship between senior-friendly destination attributes and travel intention. Implications and suggestions are provided for tourism researchers, destination managers, and travel service providers.

Exploring physical activity in Central Chile during the Early Ceramic Period and Late Intermediate Period (200-1450 CE).

We explore physical activity in early farming societies of Central Chile during the Early Ceramic and Late Intermediate Periods (200-1450 CE), a time of technological changes and intensification of food production. The existence of differences in entheseal changes (EC) between females and males is evaluated in two periods with different subsistence strategies. EC were recorded with method in the upper and lower limbs of 56 male and female adult individuals from the Early Ceramic Period (ECP) and Late Intermediate Period (LIP). Hierarchical multiple factor analysis and hierarchical clustering on principal components were performed using an exploratory approach. In all the analyzed limbs, the EC scores are higher in males than females, which may be explained by sexual dimorphism. A constant overlap in the variability of the EC among males and females is observed. However, the EC scores show a different pattern of variability when comparing ECP females with LIP females. The results show that the scores increase in the latter. In contrast, when comparing ECP males with LIP males scores decrease in the later period. Physical activity in the societies of Central Chile with incipient agriculture did not vary by sex or during the timeframe studied. Nevertheless, the interaction between sex-period and the variability pattern of the EC from one period to another suggests different ways of doing similar physical activities or different emphases for similar biomechanical actions.

Comparison of Three Approaches to Assess the Flavour Characteristics of Scotch Whisky Spirit

This study compared the use of three sensory and analytical techniques: Quantitative Descriptive Analysis (QDA), Napping, and Gas Chromatography-Mass Spectrometry (GC-MS) for the assessment of flavour in nine unmatured whisky spirits produced using different yeasts. Hierarchical Multiple Factor Analysis (HMFA) showed a similar pattern of sample discrimination (RV scores: 0.895–0.927) across the techniques: spirits were mostly separated by their Alcohol by Volume (ABV). Low ABV spirits tended to have heavier flavour characteristics (feinty, cereal, sour, oily, sulphury) than high ABV spirits, which were lighter in character (fruity, sweet, floral, solventy, soapy). QDA differentiated best between low ABV spirits and GC-MS between high ABV spirits, with Napping having the lowest resolution. QDA was time-consuming but provided quantitative flavour profiles of each spirit that could be readily compared. Napping, although quicker, gave an overview of the flavour differences of the spirits, while GC-MS provided semi-quantitative ratios of 96 flavour compounds for differentiating between spirits. Ester, arenes and certain alcohols were found in higher concentrations in high ABV spirits and other alcohols and aldehydes in low ABV spirits. The most comprehensive insights on spirit flavour differences produced by different yeast strains are obtained through the application of a combination of approaches.

Innovation and the entrepreneurial performance in women-owned small and medium-sized enterprises in Pakistan

PurposeThis study is an attempt to examine the impact of entrepreneurship and innovation on the entrepreneurial performance of women-owned small and medium-sized enterprises in Pakistan. The study also investigates the mediating role of innovation in the relationship between entrepreneurship and performance in women-owned businesses from emerging economy of Pakistan. Methodology/ApproachData was collected from the 261 registered women entrepreneurs from all the fifteen Pakistani women's chambers of commerce and industries by employing self-administered questionnaires for complementing a comprehensive literature review on innovation and women's entrepreneurship and performance. Data were analyzed using hierarchical multiple regression and factor analysis. FindingsThe results show a significant direct relationship between entrepreneurship, innovation and the entrepreneurial performance. The two entrepreneurial traits, namely risk taking tendency and the need for achievement have a significant effect on innovation and the entrepreneurial performance. Moreover, innovation has a direct relationship with performance and also it mediates the entrepreneurship- performance link of women- owned small and medium-sized enterprises in Pakistan. Research limitationsThe present study is limited to only women entrepreneurs who are registered with the chambers of commerce and industries in Pakistan. Women entrepreneurs associated with federations and several women business development centers have not been considered in the population. Practical implications: This study emphasizes the significance of entrepreneurship to innovation and the entrepreneurial performance in women-owned SMEs. It is a major contribution to understanding the existing practices and theories of entrepreneurship and innovation in enterprises. In order to improve performance and innovation, management should focus and promote risk taking and the need for achievement. The government should initiate support programs for women's entrepreneurial activities; promote fair competition between men and women and give special grants to women entrepreneurs in order to improve their performance.

Social hierarchy of pain and its connection to the memory of previously suffered pain.

BackgroundPain is a perception conditioned both by the painful experience and by each society’s collective imagination. The general objective of the project which this work forms part of it was to discover what citizens think about different aspects of this complex experience. More precisely, this paper’s objective is to get to know which is the worst pain that can be suffered according to Spaniards and what determines that hierarchy, bearing in mind that this work has chosen a broad definition of pain, including pains of different origins, namely, physical, psychological, and emotional pain.Materials and methodsThe data from the CIS 3137 study “Social perceptions of pain” have been used, which is a survey module designed by the Institute of Advanced Social Studies (IESA) of the Spanish National Research Council (CSIC). A hierarchical multiple factor analysis has been performed, using the SPSS statistical analysis software, where the dependent variable is the citizen’s opinion on which is the worst pain that can be suffered, recoded according to the origin of pain (physical, psychological, and emotional pain). Sociodemographic variables and variables linked to the experience of pain have been included as independent variables.Results and conclusionAlthough the most frequent pains among Spanish citizens are those of a physical origin, especially those linked to musculoskeletal problems and pains of an orofacial origin, when they are asked about the worst pain a person can suffer, they do not mention this type of pain, but those of an emotional origin. It has also been possible to confirm that the pain that citizens refer to when asked about the worst pain that can be suffered, and, therefore, the hierarchy of pain held by Spanish citizens as a group, is conditioned, although not determined, by the pain that has previously been suffered – by one’s own experience of pain.

Urban and rural river restoration in France: a typology

AbstractRiver restoration (RR) is widely practiced in both rural and urban contexts by combining various goals and measures. The theoretical discourse on RR not yet adequately reflects this breadth of restoration practice. In this study, we investigated 110 French RR projects implemented between 1980 and 2015. We analyzed projects considering eight key design features, main project motivation, restoration goals, project dates, costs, size, funding, river annual discharge, and implemented evaluation procedures. The study (1) provides a detailed account of the French RR effort, (2) compares restoration efforts in urban and rural contexts, and (3) establishes a RR project typology. The results also show that urban RR comprises a wider range of goals and measures than its rural counterpart, includes restoration of riparian habitats, and integrates ecological and social goals. A hierarchical multiple factor analysis yielded five types of projects, Fish RR (14% of the urban and 53% of the rural projects), Blue RR (4%, 7%), Water Framework Directive RR (36%, 40%), Flood protection RR (14%, 0%), and Human RR (32%, 0%). We suggest that the restoration community needs databases that use a project typology as developed in this study. This approach would take into account the multiple facets of RR projects, enabling more transparency into their communication and allow more suitable project comparisons.

Sensory profiles of artisan goat cheeses as influenced by the cultural context and the type of panel

SummaryThe research was performed to evaluate influence of the cultural context and the type of panel on sensory profiles of artisan goat cheeses. Two types of sensory panels from the goat cheese‐making region (experienced/trained artisan cheese producers vs. goat cheese consumers) and two types of sensory panels from a city area (trained descriptive panellists vs. cheese consumers) were formed. The sensory profiles generated with QDA® and Flash Profile techniques were compared using the hierarchical multiple factor analysis at two levels: type of panel and geographical area. This study demonstrated that sensory panels from the goat cheese‐making region used lesser attributes for characterising cheeses. Some discrepancies in sensory profiles were found between the two consumer panels. The two experienced/trained panels were more discriminative and their sensory profiles were similar. Recruiting artisan goat cheese producers for sensory profiling of their own products is a reasonable alternative when training is appropriately conducted.

Perception and Description of Premium Beers by Panels with Different Degrees of Product Expertise

The present study compares subjects with varying degrees of product expertise with regards to their ability to provide a sensory profile of beverages. Eight premium beers were evaluated by three different panels using a Napping® test, followed by a descriptive task. Two panels were constituted of consumers, classified according to their self-assessed product expertise into “Novices” (N = 14) and “Enthusiasts” (N = 26). The sensory panel at a large brewery, and a group of master brewers constituted the third panel (“Experts”, N = 15). The Napping® data from the three panels were digitalized using a coordinate system, whereas attributes were entered separately and treated as frequency table crossing products and attributes. The position data were analyzed by Hierarchical Multiple Factor Analysis (HMFA). Partial Least Squares-Discriminant Analysis (PLS-DA) was used to test differences between the three panels with regards to the use of attributes. The HMFA results showed a separation of the samples into two distinct groups on the first dimension, whereas the second dimension highlighted the specificity of two of the samples. RV coefficients between partial configurations obtained from the three panels were all above 0.90, indicating high configurational similarity. In contrast, PLS-DA showed significant differences in the use of attributes, particularly between Experts and Novices, suggesting that product expertise is more associated with descriptive, rather than perceptual, ability.

Sensory Perception of Yackwa (Korean Traditional Fried Cookie) by Consumer Groups of Different Age Using the Sorted Napping Procedure

AbstractThis study was conducted to investigate sensory perception of Korean traditional fried cookie (Yackwa) by two Korean consumer groups of different age (ages in 20s and ages in 30–40s). The sorted napping procedure was applied to evaluate a set of 12 commercial Yackwa samples. Each subject was asked to locate each sample based on perceived similarities and then group the samples according to her own criteria. Verbalization task for each group of samples was also carried out. Data obtained from the sorted napping test were subjected to hierarchical multiple factor analysis. Both groups differentiated the samples mainly according to textural properties resulting from their manufacturing type. However, the older group located the samples from the same company that produces Yackwa following traditional recipe close to each other even though their types were different, whereas the younger group sorted the samples mainly by the manufacturing type. This difference in perception between two age groups might have resulted from the fact that older consumers recognized the flavor of more traditional products better than the younger group due to more exposure to those products.Practical ApplicationsThis study investigated consumer perception of Yackwa (traditional Korean cookie). It is expected that more consumer‐based characterization, which is explained by their own words, can help researchers to understand what important sensory dimensions are for consumers. In addition, by comparing different consumer groups, this study will provide insight that can help to develop effective strategies for commercializing traditional foods for targeted generations of consumers.

Comparison of perceptual representations of soymilk products between consumers with different nuttiness rating tendencies

Internal perceptual representations of soymilk between 2 groups with different nuttiness rating tendencies were compared. A total of 30 consumers who emphasized a cooked soybean flavor (SOY group) and 30 consumers who emphasized sweetness and nut flavors (SWTNUT group) for rating nuttiness flavor participated. A sorted napping procedure was applied to 10 commercial soymilk samples. Data were analyzed using hierarchical multiple factor analysis. Both groups differentiated samples similarly, primarily on the basis of sweetness and textual properties. Noticeable differences between the 2 groups were found in the use of verbal descriptors. Sensory terms were more frequently elicited from the SOY group, whereas affective terms were more frequently elicited from the SWTNUT group. There was no difference in sensory perceptions between the 2 groups of consumers, but each group used different strategies to verbalize sensory perceptions.

Analysis of the Slovak consumer behaviour regarding the organic food purchase

The aim of this work was to determine whether Slovak consumers buy organic food, to analyze the frequency of organic food shopping and to examine the monthly expenditure of organic food. Questionnaire survey involved 271 respondents. The hierarchical multiple factor analysis was used for the segregation and classification of consumers into representative groups. The group of respondents was based on algorithms divided into three groups. The first group consists of those respondents who do not buy organic food. In the second group of respondents, prevalent was occasional purchase of organic food. Prevailing monthly expenditure on buying organic food is the amount of 10–15 euros. The third group of respondents does weekly shopping of organic foods, these foods make costs mostly 15 to 20 euros per month.

Quality and availability of organic foods by Slovak consumers

The increasing consumer demand for organic products caused that the organic food market has expanded in all continents of the world. Organic foods represent a specific segment of the food market. Currently land area farmed organically in Slovakia represents 9% of the total agricultural land. In this work we identified organic foods purchase by Slovak consumers, the availability, reasons of purchase and quality assortment of organic foods at the Slovak market. Questionnaire survey involved 271 respondents. The Hierarchical multiple factor analysis was used for the segregation and classification of consumers into representative groups. The group of respondents was based on algorithms divided into three groups. In the first group of respondents, prevalent are responses that assortment is not sufficient and no answer, in the second group think that organic food assortment is not sufficient, and in the third group of respondents also dominates opinion that is not sufficient. At the question of organic food quality in all three groups is prevalent opinion that it is rather high, in the first group nearly the third of respondents considered the quality of organic foods as rather low, in the second group of respondents is rate: „rather low“ response and „rather high“ almost equal. In the third group of respondents strongly dominated response that the quality of organic food is rather high. Regarding the availability of organic products at the Slovak market, 16% of respondents considered it to be sufficient, 54% of consumers considered assortment as not enough available for all. We also analyzed the reasons of buying organic food. 42% of respondents reported that the main reason for buying organic food is a concern for the environment and landscape, 33% of respondents state it is a pleasure and the opportunity to try something unusual, 11% reported confidence in the quality of organic food and 7% their health care. Environmental education in the family since childhood and the opinions of friends significantly influence consumers when buying food and organic food. We assume that, organic food will be processed minimally in the future. Consumer demand for organic food we recommend to increase by increased support of the direct sales, informing about quality of organic foods and their beneficial effects on the human body at basic and secondary schools, organizing of excursions to organic farms and processing companies.

PrefMFA, a solution taking the best of both internal and external preference mapping techniques

In product optimization, preference mapping techniques are widely used. Although their advantages and benefits are evident, these methods also show some limitations. For example, in external preference mapping (PrefMap), in general only two dimensions are used in the individual regression models, and there is no evidence that these dimensions are relevant for liking. Moreover, potentially important information which could explain the liking scores and which could be present on the third and higher dimensions of the sensory space is often not considered.For that reason, a new methodology, called PrefMFA, is proposed. This methodology based on Multiple Factor Analysis (MFA) is in between internal and external preference mapping since it takes the dimensions from the “common” product space between external (usually sensory) and the hedonic scores in the individual regressions. An extension to Hierarchical Multiple Factor Analysis (HMFA), when more than one external matrix is available, is also proposed.The PrefMFA methodology is applied to two datasets according to two different strategies of analyses, and the advantages of PrefMFA over PrefMap are shown. More precisely, the help for the interpretation as well as the various criteria proposed by MFA (such as the partial axes representation, the Lg, and the RV coefficients) help to better understand the strength of the underlying relationship between the external information and liking.

Structural relationships among vegetation, soil fauna and humus form in a subalpine forest ecosystem: a Hierarchical Multiple Factor Analysis (HMFA)

Aboveground vegetation, four belowground fauna groups and humus composition have been analyzed in order to investigate the links between autotrophic and heterotrophic communities in a Norway-spruce mountain forest in Tours-en-Savoie (France). The aboveground plant community was recorded in small patches corresponding to contrasting microhabitats. Animal communities and humus layers were sampled within the same patches. The relationships between humus profile, faunistic and floristic compositional gradients were investigated by Multiple Factor Analysis (MFA) and, for the first time in ecology, a Hierarchical Multiple Factor Analysis (HMFA) was used to interpret differences among humus layers. The analysis revealed a pattern with three main groups of microhabitats. The thorough study of separate humus layers could explain this result. The interplay of plant-animal-soil interactions is likely to drive the ecosystem toward three alternative states supporting humus traditional classification between mull-mor-moder. HMFA revealed the importance of depth to explain this contrast among humus forms, using humus layers as diagnostic tools in both inert and living components. HMFA also showed contrast between unexploited and exploited parts of the forest, but the study of soil and vegetation indicate that this contrast does not only hold in forest management but also in geomorphology. RV-coefficients among the six groups of variables showed significant fauna-fauna relationships in almost all humus layers except Actinedida. Plant-soil interactions are not as strong as expected and are even weaker when the soil in question is deep. In addition, HMFA failed to show direct interactions between plant and soil fauna but, paradoxically, HMFA does suggest that indirect plant-fauna interactions are at the focus of the ecosystem strategy that leads to the differentiation of ecological niches within the forest mosaic.

Establishment of a cassava (Manihot esculenta Crantz) core collection based on agro-morphological descriptors

International Institute of Tropical Agriculture maintains 2544 cassava accessions (Manihot esculenta Crantz) from 28 countries in its field bank. Being vegetatively propagated, this poses challenges in maintenance in terms of cost as well as in labour requirements. A core collection representing the range of phenotypic diversity present in the entire collection would enhance the conservation aspects and increase the potential for its exploitation in crop improvement programmes. The present study aimed to establish a core collection using 40 agro-morphological traits evaluated at two locations using a different number of accessions in each location. To meet the challenges generated by the types of variables and include maximum diversity in the core collection, a sequential strategy based on five major concepts was used: hierarchical multiple factor analysis allowing the mixture of variables of different kinds; three-way analysis that included the effect of genotype × environment interaction in the clustering process; linear discriminant function to assign all those individuals who were included in one location but not in the other to the groups that were generated from the common number of accessions evaluated in both locations; and D-allocation method to select samples from each cluster. The representativeness of the core subset to the entire collection was further estimated by comparing means and variances, range, and distances between accessions. The established cassava core collection consisted of 428 accessions that conserved 15% higher phenotypic diversity with no redundancies. The phenotypic diversity represented in this core collection will be a guide to users of cassava germplasm in their crop improvement programmes.