Multicollinearity Diagnosis Research Articles

Nonlinear seismic site response classification using K-means clustering algorithm: Case study of the September 6, 2018 Mw6.6 Hokkaido Iburi-Tobu earthquake, Japan

Spectral ratio methods have been widely used in evaluation of nonlinear seismic site response. Nevertheless, it remains inefficient and subjective to identify stations with nonlinear site response according to empirical threshold values of spectral ratio nonlinear degree indicators. This study, which was the first to apply the machine learning clustering algorithm to address this problem, used the September 6, 2018 Mw6.6 Hokkaido Iburi-Tobu earthquake (Japan) as an example. First, we calculated the surface/borehole and horizontal/vertical spectral ratios using strong ground motion data recorded by KiK-net vertical array and K-NET stations, respectively. The degree of nonlinear site response (DNL) and percentage of nonlinear site response (PNL) were computed using the difference between the strong motion of the mainshock and weak aftershocks as the reference for linear site response. Then, the K-means clustering algorithm was incorporated in the identification of nonlinear site response using the DNL, PNL, strength of ground motion (PGA) and site condition (VS20 or VS30) as explanatory variables. After careful multicollinearity diagnosis and confirmation of the optimum clustering number, we successfully classified the stations into two clusters with nonlinear and linear site responses. Overall, the clustering results were found in good agreement with the classification results based on empirical thresholds of several nonlinear indicators. For the stations identified with nonlinear site response, the reduction of amplification and frequency shift could be observed from the spectral ratio curves regarding the ground motions in the mainshock and the reference weak aftershocks, demonstrating typical nonlinearity response characteristics. Furthermore, a comprehensive indicator of nonlinear site response occurrence probability (NLscore) was obtained from a linear weighted combination of the normalized variables (PGA, VS30/VS20, DNL and PNL). The NLscore ranking of the top several stations was found consistent with the clustering identification results, irrespective of the choice of combination scheme. It was demonstrated that the performance of clustering algorithm in this application was satisfactory and that the identification results were convincing and robust. This work provides an enlightening example of using state-of-art machine learning technique to solve the traditional earthquake engineering problems.

Multivariate analysis reveals key traits of fall armyworm resistance in tropical popcorn genotypes

ABSTRACT In the present study it is hypothesized that the germplasm of popcorn of tropical regions shows resistance to FAW, and the multivariate analysis can characterize the main traits of this resistance. Thus, the aim of this study was to identify key traits used to select fall armyworm resistance in popcorn by using factor and cluster analyses. Sixteen biological traits were evaluated at larval and pupae stages: total number of armyworm, time larval period, final mass of caterpillars, total mass of caterpillars, mean mass of caterpillars, pupae period, mass of pupae and some indices used to evaluate the food consumed by the caterpillars. Multicollinearity diagnosis, factor and canonical analysis, and the genetic divergence among the genotypes were performed to implement multivariate analysis. The groups were established according to Mojena (1977). After multicollinearity test, only five traits were retained for further analysis. The factor analysis divided the five traits into two factors: the first factor included time larval period, metabolized food and mass of pupae; the second was composed of total number of armyworm and stool mass. Unweighted Pair Group Method with Arithmetic Mean (UPGMA) grouped the eighteen genotypes in three clusters. The present study provides insights of popcorn resistance to fall armyworm, and the multivariate analytical approach used here is directly applicable to any species and set of traits exhibiting correlation.

Implicações da análise univariada e multivariada na dissimilaridade de acessos de feijão comum

Univariate and multivariate analyses have been routinely used to discriminate genotypes in breeding programs, but their relationship is not always considered. The objective of this study was to relate univariate and multivariate analyses on the dissimilarity among common bean accessions based on characteristics such as days until emergence, plant height, first pod insertion height, harvest days, final stand, number of pods per plant, pod length, 100-grain weight, grain yield and seed length, width and thickness. Data were submitted to univariate analysis of variance, with a comparison of means by the Scott-Knott test, and multivariate analysis to estimate the genetic divergence based on the generalized Mahalanobis distance, grouping the accesses through Tocher’s method. The relative importance of the characters was also estimated by Singh's method, performing the multicollinearity diagnosis. There was a significant difference for all the evaluated characteristics, showing the existence of variability between the accesses. The greater variability found was related to seed characteristics. It was possible to relate univariate and multivariate analyses, since there was a relationship between the groups generated by Scott-Knott’s test and the percentage of the relative contribution of the evaluated characters. However, the clustering obtained in the univariate analysis for the characteristic that contributed the most to the divergence between the accessions was not compatible with the cluster generated by Tocher’s method. The coupled use of these analyses can help breeders to decide, choosing superior genotypes and that present wide genetic divergence, mainly for the characteristics of greater interest.

Kinematic Calibration Based on the Multicollinearity Diagnosis of a 6-DOF Polishing Hybrid Robot Using a Laser Tracker

A general methodology for ensuring the geometric accuracy of a 6-DOF polishing hybrid robot having a 3UPS and UP parallel mechanism and a 3-DOF wrist is presented. The process is implemented in three steps: formulation of the error model containing complete source errors such as offset errors of the actuated joints and structural errors of the joints and links utilizing product of exponentials formula and screw theory. Measurement of the full pose error twist with a specially designed measurement tool having three reference points was undertaken. Identification of the source errors by a stepwise identification strategy to overcome the ill-conditioned problem arising from the multicollinearity and development of a linearized error compensator was completed. An experiment has been carried out on the prototype, and the results show that, after calibration, a position volumetric error of 0.07 mm and an orientation error of 0.07 degrees can be achieved over the cubic task workspace with repeatability of 0.016 mm and 0.010 degrees.

Interference of sample size on multicollinearity diagnosis in path analysis

Abstract: The objective of this work was to evaluate the interference of sample size on multicollinearity diagnosis in path analysis. From the analyses of productive traits of cherry tomato, two Pearson correlation matrices were obtained, one with severe multicollinearity and the other with weak multicollinearity. Sixty-six sample sizes were designed, and from the amplitude of the bootstrap confidence interval, it was observed that sample size interfered on multicollinearity diagnosis. When sample size was small, the imprecision of the diagnostic criteria estimates interfered with multicollinearity diagnosis in the matrix with weak multicollinearity.

Research Article Correlation among traits as criterion of cotton genotypes indirect selection

The cotton crop has notable importance for Brazilian and worldwide economy. Relationships knowledge among characters allows developing alternatives that increase the genetic progress with simultaneous selection of several traits. The use of correlation estimates for use in indirect selection helps the breeder in the selection of characters and facilitates progress in the selection. This work evaluates the phenotypic correlation and unfolding of the direct and indirect effects of morphological and technological fiber traits as criterion of indirect selection in cotton genotypes. The experiment was carried out at Fazenda Capim Branco, a research station of Universidade Federal de Uberlândia, 2014/2015 season. 21 genotypes were evaluated, 19 commercial genotypes F3 proceeding from the Programa de Melhoramento Genetico Algodoeiro (PROMALG) and two commercial control: FM 966 and DP 555. The experimental design was the randomized complete block with four replications. The experimental parcel was made of four rows of 5 meters cotton plants length, spaced of 0.9 meters. The morphological and agronomic traits evaluated were: a) chlorophyll content (TCA and TCB); b) number of fruiting branches (NFB); c) 1st fruit node height (NH); d) first, second and third-position bolls percentage (FPB, SPB, TPB); e) leaf Area Index (LAI); f) cotton seed yield (CSP); g) lint yield (LP); and h) Lint yield (LY). The fiber quality traits evaluated with HVI (High Instrument Volume): i) Fiber Length average (UHML); ii) Uniformity of length (UI); iii) Short fiber Index (SFI); iv) Fiber Strengh (STR); v) Elongation (ELG); vi) Micronaire (MIC), and vii) Fiber Maturity (MR). Statistical analyses and multicollinearity diagnosis were carried out with the software GENES. The intrinsic traits of fiber quality, according to correlations, influenced cotton seed yield and also morphological traits. Based on phenotypic correlations and according to path analysis, the higher micronaire index of favorable effect on cotton seed yield was identified due to a positive correlation and direct effect. Thus, micronaire index can be an indirect selection criterion of cotton genotypes.

Multicollinearity in canonical correlation analysis in maize.

The objective of this study was to evaluate the effects of multicollinearity under two methods of canonical correlation analysis (with and without elimination of variables) in maize (Zea mays L.) crop. Seventy-six maize genotypes were evaluated in three experiments, conducted in a randomized block design with three replications, during the 2009/2010 crop season. Eleven agronomic variables (number of days from sowing until female flowering, number of days from sowing until male flowering, plant height, ear insertion height, ear placement, number of plants, number of ears, ear index, ear weight, grain yield, and one thousand grain weight), 12 protein-nutritional variables (crude protein, lysine, methionine, cysteine, threonine, tryptophan, valine, isoleucine, leucine, phenylalanine, histidine, and arginine), and 6 energetic-nutritional variables (apparent metabolizable energy, apparent metabolizable energy corrected for nitrogen, ether extract, crude fiber, starch, and amylose) were measured. A phenotypic correlation matrix was first generated among the 29 variables for each of the experiments. A multicollinearity diagnosis was later performed within each group of variables using methodologies such as variance inflation factor and condition number. Canonical correlation analysis was then performed, with and without the elimination of variables, among groups of agronomic and protein-nutritional, and agronomic and energetic-nutritional variables. The canonical correlation analysis in the presence of multicollinearity (without elimination of variables) overestimates the variability of canonical coefficients. The elimination of variables is an efficient method to circumvent multicollinearity in canonical correlation analysis.

Linear associations among phenological, morphological, productive, and energetic-nutritional traits in corn

Abstract: The objective of this work was to verify if there is linear dependence between the phenological, morphological, and productive traits and the energetic-nutritional ones in early maturing and super-early maturing corn genotypes. A total of 36 early maturing and 22 super-early maturing corn genotypes were evaluated in a randomized complete block design with three replicates, and the phenological, morphological, productive, and energetic-nutritional traits were measured. The matrix of phenotypic correlation coefficients among traits was determined; the multicollinearity diagnosis was carried out within each group of traits; and the canonical correlation analysis was performed. Linear dependence was observed between the groups of phenological, morphological, and productive traits and of the energetic-nutritional ones. In early maturing genotypes, significant canonical correlation shows the existence of linear dependence between the morphological and energetic-nutritional traits. The significant canonical pair shows that taller plants have lower amylose contents in the grains and reduced nitrogen-corrected apparent metabolizable energy. In super-early maturing genotypes, significant canonical correlations of the phenological, morphological, and productive traits with the energetic-nutritional ones indicate that a greater number of days from sowing to female flowering and ear insertion height, as well as a lower number of ears, increase ether extract contents in the grains.

Relationship between fruit traits and contents of ascorbic acid and carotenoids in peach

Abstract: This study aimed to evaluate the relationship between fruit traits and their direct and indirect effects on the content of ascorbic acid and ca - rotenoids in peaches and nectarines. The traits fruit mass (FM); equatorial diameter (ED); suture diameter (SD); polar diameter (PD); pulp firmness (FIR); soluble solids (SS); titratable acidity (TA); SS/TA ratio; contents of ascorbic acid (AA) and carotenoids (CT); and skin and pulp color were evaluated in 28 peach cultivars, and two nectarine cultivars. The phenotypic correlation coefficients were estimated (rf), and after multicollinearity diagnosis, unfolding was carried out in direct and indirect effects of the explanatory variables in the response variable by using path analysis. The strongest correlations were found between FM, SD, ED, and PD, and between carotenoid content and °h pulp. The traits considered in the path diagrams are not the main determinants of the ascorbic acid content. The yellow color of the pulp has the potential for indirect selection for carotenoid content.

Relationship between morphological traits and body mass of Miichthys miiuy

Miichthys miiuy Basilewsky is a commercially important marine fish in China. To understand the relationship between body mass and morphometric traits of M. miiuy, body mass (Y, g) and other eight morphometric traits (cm), e.g., total length (X1), body length (X2), head length (X3), trunk length (X4), tail length (X5), tail handle length (X6), tail handle height (X7), and body height (X8) of 524 individuals of about 12-month-old M. miiuy were measured in this study. The methods of correlation analysis, multiple regression analysis, and path ana-lysis were used to analyze the data. Also, the best model reflecting the relationship between morphometric traits and body mass were obtained through model fitting screening. The results showed that the correlation coefficients between every two traits ranged from 0.834 to 0.941, and all reached significance level (P<0.01). And body length was ruled out from the multiple regression analysis by the method of multicollinearity diagnosis. We established a regression equation using the rest morphologi-cal traits as variables and body mass as dependent variable, in which direct and indirect path coefficients and determinant coefficients were calculated. Four morphometric traits achieved highly significant levels (P<0.01), and they were the main decisive morphometric traits linked to body mass. The multiple linear regression model equation was established as Y＝-444.188 +7.943X1+12.861X4 +38.254X7 + 42.722X8. Among them, the trait with the strongest direct effect on body mass was the body height (0.351), then the total length (0.335). In addition, the best model fitted for the individuals was y=0.013 x2.891 for total length, y=2.028 x2.751 for body height. The method of model fit test showed the accuracy of total length estimated body mass was higher than that of body height. This research may provide an important reference for the effective use of morphological traits for breeding of M. miiuy.

Linear relations among phenological, morphological, productive and protein-nutritional traits in early maturing and super-early maturing maize genotypes

The objective of this study was to verify through canonical correlation analysis if there is linear dependence among phenological, morphological, and productive traits with protein-nutritional traits in maize genotypes. The experiments were carried out in a randomized block design with three repetitions. In one experiment, 36 early maturing maize genotypes were evaluated and in another experiment 22 super-early maturing maize genotypes. The canonical groups were determined and structural equation models were elaborated. The matrix of phenotypic correlation coefficients among 23 traits was determined. Thereafter, the multicollinearity diagnosis was conducted within each group of traits. The canonical correlation analysis was performed within the groups: phenological versus protein-nutritional, morphological versus protein-nutritional, and productive versus protein-nutritional. In early maturing maize genotypes, the canonical correlations were not significant among groups of traits showing that the traits cannot be used as indicative of protein-nutritional quality in the indirect selection of plants. Moreover, in super-early maturing maize genotypes, the significant canonical correlation among phenological versus protein-nutritional traits and among productive versus protein-nutritional indicate that the traits number of days from sowing to female flowering, number of ears, and grain yield can be used for indirect selection as indicative of protein nutritional quality in maize grains.

Physiological characteristics in seeds of the common bean under multicollinearity and conditions of salinity

The objective of this work was evaluate multicollinearity effect and discard the variables which are based on multicollinearity reduction in diversity analysis of common bean genotypes, related to seeds physiological quality, in different salinity levels in germination substrate. The common bean seed germination test for six cultivars and seven landrace genotypes was performed in paper rolls (germitest), imbibed in NaCl solutions on the osmotic potentials of 0.0; -0.3 and -0.6 MPa, maintained in germinated Mangelsdorff type at temperature of 25 °C, on constant light. The experimental design was completely randomized in factorial arrangement 13 x 3 (genotype x osmotic potential), with four replications with 25 seeds, totaling 100 seeds per treatment. The carried out evaluations were: weight of a thousand seeds, germination mean time, primary root protrusion in five and nine days after seeding, normal seedling percentage, hypocotyl and primary root length and dry matter from aerial part and roots. The multicollinearity diagnosis was carried out on the phenotypic correlation matrix and the characteristic discard was preceded through the canonical variable technique. To evaluate the multicollinearity effect, the Tocher cluster method was used before and after the variables discard. The proposed discard methodology of variables is efficient on the multicollinearity reduction and the number of discarded physiological quality descriptors is higher on the potentials of -0.3 and -0.6 MPa, under salt stress conditions, need to be discarded three characteristics and in the absence of stress only two discarded, to became a weak condition of multicollinearity to follow with the cluster analysis. The common bean genotypes clusters are different on the severe and weak multicollinearity only under salt stress condition.

Altitude and geographic coordinates to estimate monthly rainfall in the state of Mato Grosso do Sul

Adjustment of multiple linear regression equations has allowed estimating the value of a certain climatological variable according to geographical coordinates with acceptable degree of accuracy. The aim of this study was to verify if the average monthly rainfall could be estimated according to the altitude, latitude and longitude in Mato Grosso do Sul State (MS). Rainfall data of 32 stations of MS were collected from 1954 to 2013. It were formed 384 time series (12 months × 32 sites), with different numbers of years of observations in each series. On each of the 384 monthly rainfall time series it was calculated the average (a), at least 30 years of observation, forming 12 matrices 32 x 4 (32 sites x 4 variables: altitude, latitude, longitude and monthly rainfall). It was estimated for each matrix the Pearson’s linear correlation coefficient among the variables, performing the multicollinearity diagnosis for each matrix. Correlations were unfolded by path analysis in direct and indirect effects and in each month it was used the multiple linear regression model. The altitude and latitude have greater effect on the spatial distribution of rainfall in MS. The multiple linear regression equations generated in this study will subsidize researches of crop zoning, indication for sowing times, irrigation, determination of yield potential, climate risks zoning and credit and agricultural insurance.

RELAÇÕES ENTRE CARACTERES DE PIMENTÃO PROVENIENTES DE SEMENTES COM DIFERENTES NÍVEIS DE QUALIDADE FISIOLÓGICA

RESUMO: O objetivo deste trabalho foi identificar as relações entre as variáveis morfológicas e produtivas das plantas de pimentão provenientes de sementes com diferentes níveis de qualidade fisiológica. Os experimentos foram conduzidos em ambiente protegido na Universidade Federal de Santa Maria (UFSM), no delineamento blocos ao acaso, com oito repetições. Os tratamentos consistiram de três lotes de sementes do híbrido Tiberius e um lote da cultivar Rubi Giant nas estações de cultivo primavera-verão e outono-inverno, com diferentes níveis de qualidade fisiológica com base nos testes de germinação e vigor. Foram mensurados os seguintes caracteres (variáveis): altura de planta; altura da primeira bifurcação; diâmetro do caule; diâmetro do caule abaixo da primeira bifurcação aos 50, 70 e 90 dias após o transplantio; peso médio de frutos; produção média de frutos por colheita; produção total; comprimento médio de frutos; e largura média de frutos. A seguir, para cada nível de vigor, foi estimada a matriz de correlação de Pearson, o diagnóstico da multicolinearidade e, posteriormente, a análise de trilha, considerando a produção total de frutos como variável principal. As plantas obtidas a partir de sementes de vigor alto, médio e baixo, com maior diâmetro da primeira bifurcação aos 50 e 90 dias após o transplantio, são mais produtivas na estação de cultivo primavera-verão. No outono-inverno foram verificados coeficientes de correlação e efeitos diretos de baixa magnitude para todos os níveis de vigor, indicando fraca relação de causa e efeito das variáveis explicativas mensuradas sobre a produção total de frutos de pimentão.

Análise de trilha da contaminação por aflatoxinas em grãos de milho armazenados

O objetivo deste trabalho foi determinar a magnitude dos efeitos diretos e indiretos de parâmetros físico-químicos e microbiológicos sobre a ocorrência de aflatoxinas em grãos de milho armazenados, por meio da análise de trilha. Os dados foram coletados em dois experimentos: um cultivo de inverno (colheita do milho em julho de 2012) e um cultivo de verão (colheita em fevereiro de 2013). Os grãos foram armazenados por 12 meses, em dois tipos de acondicionamentos: em sacarias convencionais de polipropileno e a granel em silo metálico. Avaliaram-se 14 características dos grãos. Calculou-se a matriz de correlações de Pearson entre todos os parâmetros, realizou-se o diagnóstico de multicolinearidade e, posteriormente, aplicou-se a análise de trilha com regressão em crista. Observou-se forte influência das condições ambientais entre as safras de inverno e de verão, e entre época e tipo de armazenagem, nos coeficientes de análise de trilha. As variáveis teor de umidade, conteúdo de cinzas, proteínas, lipídios, grãos sem defeitos, massa volumétrica, incidência de Penicillium sp., Aspergillus sp. e Fusarium sp. devem ser consideradas conjuntamente na definição da possibilidade de ocorrência de aflatoxinas na massa de grãos armazenados.

Comparison of clustering methods for study of genetic dissimilarity in soybean genotypes

This study aimed to compare clustering methods based on the dissimilarity measures and get information on genetic diversity in twelve soybean genotypes. The experiment was conducted at the State University of Mato Grosso do Sul - Unit Aquidauna. The genotypes were grown in a randomized complete block design with four replications. The following quantitative traits were measured: Plant height, first pod insertion height, number of branches and pods, mass of hundred grains and grain yield. It were evaluated the clustering methods Ward, complete linkage, median, mean linkage within group and mean linkage between groups using as dissimilarity measures the mean standardized Euclidean distance (D) and the Mahalanobis’s generalized distance (D²). The diagnosis of multicollinearity revealed adequacy of data to the proposed study. Clustering based on standardized mean Euclidean distance is distinct from those formed based on the Mahalanobis’s generalized distance, being this measure most recommended to quantify the genetic diversity in soybean genotypes based on morphological traits because it presents higher values cophenetic correlation coefficient (CCC) for all clustering methods. The mean linkage between groups’s hierarchical method formed concordant groups for D and D², being recommended for these dissimilarity measures. According to both methods, the cross between the genotypes CD238 with SYN3358 and CD238 with Potencia can generate hybrids with high heterotic effect due to different numbers of loci in which the dominance effects are evident. Key words: Dissimilarity measures, genetic divergence, Glycine max, quantitative descriptors.

English

With the aim of to evaluate the correlation and the direct and indirect effects of morphological descriptors on oil content of castor ( Ricinus communis L.) genotypes, an experiment was conducted in the agricultural years 2010/2011 and 2011/2012 at Universidade Estadual de Mato Grosso do Sul – Unit of Aquidauana (UEMS/UUA). The experimental design used was randomized blocks with three replications. The treatments consisted of seven castor genotypes (BRS Energia, IAC-2028, IAC-Guarani, BRS Nordestina, BRS Paraguacu, IAC-80 and IAC-226). The following descriptors were evaluated: plant height (PH), stem diameter (SD), number of bunches (NB), number of fruits (NF), mass of fruits (MF), mass of hundred seeds (MHS) and oil content (OC). Initially, it was determined the following genetic parameters: genotypic variance, environmental variance, coefficient of experimental variation, coefficient of genotypic variation, heritability, environmental correlations, phenotypic correlations and genetic correlations. The phenotypic correlations were deployed by path analysis in estimates of direct and indirect effects, and preceded the diagnosis of multicollinearity in explanatory variables and independent explanatory variables on the OC (primary dependent variable). The results indicated that the direct and indirect selection of genotypes with plant height, stem diameter, number of bunches and mass of hundred seeds is promising to select genotypes with high oil content in castor.

Empirical Analysis of a Generalized Linear Multianticipative Car-Following Model in Congested Traffic Conditions

This paper empirically investigates the multianticipative car-following behavior using a multiple linear regression approach. Two important aspects of the multiple linear regression of a modified generalized linear multianticipative car-following model are addressed, namely, multicollinearity between explanatory variables and the serial correlation of time series data, which have not been taken into account by previous studies. This paper analyzes the multianticipative car-following behavior using the next generation simulation (NGSIM) trajectory data. Specifically, the following issues are discussed: reaction time estimation, multicollinearity diagnosis, serial correlation test, model selection, and parameter estimation. Preliminary results show that a driver in extremely congested traffic conditions is able to react to the stimuli of the first, second, and even third leader with respect to the relative speed and the stimuli of the first or the second leader with respect to the desired space. It is reascertained that a driver is more perceptual to speed difference than to desired space behavior in extremely congested traffic conditions. Moreover, it is observed that a driver in less congested traffic conditions only reacts to the stimuli of the first leader with respect to both the relative speed and desired distance. Therefore, it is empirically believed that the stimuli perceived by a driver may be different in different traffic conditions with respect to multianticipative car-following behavior.

A Neural Network Based on Canonical Correlation for Multicollinearity Diagnosis

We review a recent neural implementation of Canonical Correlation Analysis and show, using ideas suggested by Ridge Regression, how to make the algorithm robust. The network is shown to operate on data sets which exhibit multicollinearity. We develop a second model which not only performs as well on multicollinear data but also on general data sets. This model allows us to vary a single parameter so that the network is capable of performing Partial Least Squares regression to Canonical Correlation Analysis and every intermediate operation between the two. On multicollinear data, the parameter setting is shown to be important but on more general data no particular parameter setting is required. Finally, we develop a second penalty term which acts on such data as a smoother in that the resulting weight vectors are much smoother and more interpretable than the weights without the robustification term. We illustrate our algorithms on both artificial and real data.

Path analysis in multicollinearity for fruit traits of pepper

This study aimed to use path analysis in multicollinearity to assess the genotypic correlation coefficients and their partitions in direct and indirect effects of morphological traits of fruits on the dry matter content of pepper. The experimental plot was a randomized block design with 2 replications and 10 fruits per plot. The data were obtained from the characterization of 10 accessions (varieties) of four pepper species of the genus Capsicum. We assessed the peduncle length, fruit length, largest fruit diameter, pericarp thickness, average fruit weight and dry matter content. We carried out the multicollinearity diagnosis and the ridge path analysis to partition the genotypic correlation coefficients into direct and indirect effects, considering dry matter content as the basic variable. The largest fruit diameter showed negative genotypic correlation with the basic variable; however, it is the morphological trait that showed, in isolation, the greatest importance to explain the variation of dry matter content in fruits, and may therefore be used as an auxiliary criterion in processes of indirect selection.