Bridging the gap between different statistical approaches: An integrated framework for modelling

This paper proposes a template for modelling complex datasets that integrates traditional statistical modelling approaches with more recent advances in statistics and modelling through an exploratory framework. Our approach builds on the well-known and long standing traditional idea of 'good practice in statistics' by establishing a comprehensive framework for modelling that focuses on exploration, prediction, interpretation and reliability assessment, a relatively new idea that allows individual assessment of predictions. The integrated framework we present comprises two stages. The first involves the use of exploratory methods to help visually understand the data and identify a parsimonious set of explanatory variables. The second encompasses a two step modelling process, where the use of non-parametric methods such as decision trees and generalized additive models are promoted to identify important variables and their modelling relationship with the response before a final predictive model is considered. We focus on fitting the predictive model using parametric, non-parametric and Bayesian approaches. This paper is motivated by a medical problem where interest focuses on developing a risk stratification system for morbidity of 1,710 cardiac patients given a suite of demographic, clinical and preoperative variables. Although the methods we use are applied specifically to this case study, these methods can be applied across any field, irrespective of the type of response.

Oceanic interchange and nonequilibrium population structure in the estuarine dependent Indo-Pacific tasselfish, Polynemus sheridani

We assayed mtDNA haplotype [300 base pairs (bp) control region] geography and genealogy in the Indo-Pacific tasselfish, Polynemus sheridani from its contiguous estuarine distribution across northern Australia (n = 169). Eight estuaries were sampled from three oceanographic regions (Timor Sea, Gulf of Carpentaria and the Coral Sea) to assess the impact of Pleistocene sea level changes on the historical connectivity among P. sheridani populations. Specifically, we investigated the genetic consequences of disruption to Indian-Pacific Ocean connectivity brought about by the closure of the Torres Strait. Overall there was significant population subdivision among estuaries (F-ST = 0.161, (Phi(ST) = 0.187). Despite a linear distribution, P. sheridani did not show isolation by distance over the entire sampled range because of genetic similarity of estuaries greater than 3000 km apart. However, significant isolation by distance was detected between estuaries separated by less than 3000 km of coastline. Unlike many genetic studies of Indo-Pacific marine species, there was no evidence for an historical division between eastern and western populations. Instead, phylogeographical patterns were dominated by a starlike intraspecific phylogeny coupled with evidence for population expansion in both the Gulf of Carpentaria and the Coral Sea but not the Timor Sea. This was interpreted as evidence for recent west to east recolonization across of northern Australia following the last postglacial marine advance. We argue that although sufficient time has elapsed postcolonization for populations to approach gene flow-drift equilibrium over smaller spatial scales (< 3000 km), the signal of historical colonization persists to obscure the expected equilibrium pattern of isolation by distance over large spatial scales (> 3000 km).

Direct visualization of Ras proteins in spatially distinct cell surface microdomains

Localization of signaling complexes to specific micro-domains coordinates signal transduction at the plasma membrane. Using immunogold electron microscopy of plasma membrane sheets coupled with spatial point pattern analysis, we have visualized morphologically featureless microdomains including lipid rafts, in situ and at high resolution. We find that an inner-plasma membrane lipid raft marker displays cholesterol-dependent clustering in microdomains with a mean diameter of 44 nm that occupy 35% of the cell surface. Cross-linking an outer-leaflet raft protein results in the redistribution of inner leaflet rafts, but they retain their modular structure. Analysis of Ras microlocalization shows that inactive H-ras is distributed between lipid rafts and a cholesterol-independent micro-domain. Conversely, activated H-ras and K-ras reside predominantly in nonoverlapping, cholesterol-independent microdomains. Galectin-1 stabilizes the association of activated H-ras with these nonraft microdomains, whereas K-ras clustering is supported by farnesylation, but not geranylgeranylation. These results illustrate that the inner plasma membrane comprises a complex mosaic of discrete microdomains. Differential spatial localization within this framework can likely account for the distinct signal outputs from the highly homologous Ras proteins.

Mean anisotropy of homogeneous Gaussian random fields and anisotropic norms of linear translation-invariant operators on multidimensional integer lattices

Sensitivity of output of a linear operator to its input can be quantified in various ways. In Control Theory, the input is usually interpreted as disturbance and the output is to be minimized in some sense. In stochastic worst-case design settings, the disturbance is considered random with imprecisely known probability distribution. The prior set of probability measures can be chosen so as to quantify how far the disturbance deviates from the white-noise hypothesis of Linear Quadratic Gaussian control. Such deviation can be measured by the minimal Kullback-Leibler informational divergence from the Gaussian distributions with zero mean and scalar covariance matrices. The resulting anisotropy functional is defined for finite power random vectors. Originally, anisotropy was introduced for directionally generic random vectors as the relative entropy of the normalized vector with respect to the uniform distribution on the unit sphere. The associated a-anisotropic norm of a matrix is then its maximum root mean square or average energy gain with respect to finite power or directionally generic inputs whose anisotropy is bounded above by a���0. We give a systematic comparison of the anisotropy functionals and the associated norms. These are considered for unboundedly growing fragments of homogeneous Gaussian random fields on multidimensional integer lattice to yield mean anisotropy. Correspondingly, the anisotropic norms of finite matrices are extended to bounded linear translation invariant operators over such fields.

Rates of gene rearrangement and nucleotide substitution are correlated in the mitochondrial genomes of insects

A number of studies indicated that lineages of animals with high rates of mitochondrial (mt) gene rearrangement might have high rates of mt nucleotide substitution. We chose the hemipteroid assemblage and the Insecta to test the idea that rates of mt gene rearrangement and mt nucleotide substitution are correlated. For this purpose, we sequenced the mt genome of a lepidopsocid from the Psocoptera, the only order of hemipteroid insects for which an entire mtDNA sequence is not available. The mt genome of this lepidopsocid is circular, 16,924 bp long, and contains 37 genes and a putative control region; seven tRNA genes and a protein-coding gene in this genome have changed positions relative to the ancestral arrangement of mt genes of insects. We then compared the relative rates of nucleotide substitution among species from each of the four orders of hemipteroid insects and among the 20 insects whose mt genomes have been sequenced entirely. All comparisons among the hernipteroid insects showed that species with higher rates of gene rearrangement also had significantly higher rates of nucleotide substitution statistically than did species with lower rates of gene rearrangement. In comparisons among the 20 insects, where the mt genomes of the two species differed by more than five breakpoints, the more rearranged species always had a significantly higher rate of nucleotide substitution than the less rearranged species. However, in comparisons where the mt genomes of two species differed by five or less breakpoints, the more rearranged species did not always have a significantly higher rate of nucleotide substitution than the less rearranged species. We tested the statistical significance of the correlation between the rates of mt gene rearrangement and mt nucleotide substitution with nine pairs of insects that were phylogenetically independent from one 2 another. We found that the correlation was positive and statistically significant (R-2 = 0.73, P = 0.01; R-s = 0.67, P < 0.05). We propose that increased rates of nucleotide substitution may lead to increased rates of gene rearrangement in the mt genomes of insects.

Calculation of bound and resonance states of HO2 for nonzero total angular momentum

Bound and resonance states of HO2 have been calculated quantum mechanically by the Lanczos homogeneous filter diagonalization method [Zhang and Smith, Phys. Chem. Chem. Phys. 3, 2282 (2001); J. Chem. Phys. 115, 5751 (2001)] for nonzero total angular momentum J = 1,2,3. For lower bound states, agreement between the results in this paper and previous work is quite satisfactory; while for high lying bound states and resonances these are the first reported results. A helicity quantum number V assignment (within the helicity conserving approximation) is performed and the results indicate that for lower bound states it is possible to assign the V quantum numbers unambiguously, but for resonances it is impossible to assign the V helicity quantum numbers due to strong mixing. In fact, for the high-lying bound states, the mixing has already appeared. These results indicate that the helicity conserving approximation is not good for the resonance state calculations and exact quantum calculations are needed to accurately describe the reaction dynamics for HO2 system. Analysis of the resonance widths shows that most of the resonances are overlapping and the interferences between them lead to large fluctuations from one resonance to another. In accord with the conclusions from earlier J = 0 calculations, this indicates that the dissociation of HO2 is essentially irregular. (C) 2003 American Institute of Physics.

Transdermal penetration of vasoconstrictors - Present understanding and assessment of the human epidermal flux and retention of free bases and ion-pairs

Purpose. As reductions in dermal clearance increase the residence time of solutes in the skin and underlying tissues we compared the topical penetration of potentially useful vasoconstrictors (VCs) through human epidermis as both free bases and ion-pairs with salicylic acid (SA). Methods. We determined the in vitro epidermal flux of ephedrine, naphazoline, oxymetazoline, phenylephrine, and xylometazoline applied as saturated solutions in propylene glycol: water (1: 1) and of ephedrine, naphazoline and tetrahydrozoline as 10% solutions of 1: 1 molar ratio ion-pairs with SA in liquid paraffin. Results. As free bases, ephedrine had the highest maximal flux, Jmax = 77.4 +/- 11.7 mug/cm(2)/h, being 4-fold higher than tetrahydrozoline and xylometazoline, 6-fold higher than phenylephrine, 10-fold higher than naphazoline and 100-fold higher than oxymetazoline. Stepwise regression of solute physicochemical properties identified melting point as the most significant predictor of flux. As ion-pairs with SA, ephedrine and naphazoline had similar fluxes (11.5 +/- 2.3 and 12.0 +/- 1.6 mug/cm(2)/h respectively), whereas tetrahydrozoline was approximately 3-fold slower. Corresponding fluxes of SA from the ion-pairs were 18.6 +/- 0.6, 7.8 +/- 0.8 and 1.1 +/- 0.1 respectively. Transdermal transport of VC's is discussed. Conclusions. Epidermal retention of VCs and SA did not correspond to their molar ratio on application and confirmed that following partitioning into the stratum corneum, ion-pairs separate and further penetration is governed by individual solute characteristics.

Variabilidade gen��tica em fam��lias de meios-irm��os de cafeeiro robusta

O cultivo do caf�� �� uma das atividades do agroneg��cio de maior import��ncia socioecon��mica dentre as diferentes atividades ligadas ao com��rcio agr��cola mundial. Uma das maiores contribui����es da gen��tica quantitativa para o melhoramento gen��tico �� a possibilidade de prever ganhos gen��ticos. Quando diferentes crit��rios de sele����o s��o considerados, a predi����o de ganhos referentes a cada crit��rio tem grande import��ncia, pois indica os melhoristas sobre como utilizar o material gen��tico dispon��vel, visando obter o m��ximo de ganhos poss��vel para as caracter��sticas de interesse. O presente trabalho foi instalado em julho de 2004, na Fazenda Experimental de Bananal do Norte, conduzida pelo Incaper, no distrito de Pacotuba, munic��pio de Cachoeiro de Itapemirim, regi��o Sul do Estado, com o objetivo de selecionar as melhores plantas entre e dentro de prog��nies de meios- irm��os de Coffea canephora, por meio de diferentes crit��rios de sele����o. Foram realizadas an��lises de vari��ncia individuais e conjuntas para 26 prog��nies de meios- irm��os Coffea canephora. O delineamento experimental utilizado foi em blocos ao acaso com quatro testemunhas adicionais com quatro repeti����es e parcela composta por cinco plantas, com o espa��amento de 3,0 m x 1,2 m. Neste trabalho, considerou-se os dados das ��ltimas cinco colheitas. As caracter��sticas mensuradas foram: florescimento, matura����o, tamanho do gr��o, peso, porte, vigor, ferrugem, mancha cerc��spora, seca de ponteiros, escala geral, porcentagem de frutos boia e bicho mineiro. Todas as an��lises estat��sticas foram realizadas com o aplicativo computacional em gen��tica e estat��stica (GENES). Foram estimados os ganhos de sele����o em fun����o da porcentagem de sele����o de 20% entre e dentro, sendo as mesmas mantidas para todas as caracter��sticas. Todas as caracter��sticas foram submetidas a sele����o no sentido positivo, exceto para florescimento, porte, ferrugem, mancha cerc��spora, seca de ponteiros, porcentagem de frutos boia e bicho mineiro, para obter decr��scimo em suas m��dias originais. Os crit��rios de sele����o estudados foram: sele����o convencional entre e dentro das fam��lias, ��ndice de sele����o combinada, sele����o massal e sele����o massal estratificada. Esta disserta����o �� composta por dois cap��tulos, em que foram realizadas an��lises biom��tricas, como a obten����o de estimativas de par��metros gen��ticos. Na maioria das caracter��sticas estudadas, verificaram-se diferen��as significativas (P<0,05) para gen��tipos que, associados aos coeficientes de varia����o genot��picos e tamb��m ao coeficiente de determina����o genot��pico e �� rela����o CVg/CVe, indicam a exist��ncia de variabilidade gen��tica nos materiais gen��ticos para a maioria das caracter��sticas e condi����es favor��veis para obten����o de ganhos gen��ticos pela sele����o. Essas caracter��sticas tamb��m foram correlacionadas. Os dados foram submetidos ��s an��lises de vari��ncia e multivariada, aplicando-se a t��cnica de agrupamento e UPGMA, teste de m��dias e estudo de correla����es. Na t��cnica de agrupamento, foi utilizada a dist��ncia generalizada de Mahalanobis como medida de dissimilaridade, e na delimita����o dos grupos, o m��todo de Tocher. Foi encontrada diversidade gen��tica para as caracter��sticas associadas �� qualidade fisiol��gica, mobiliza����o de reserva das sementes, dimens��es e biomassa das pl��ntulas. Quatro grupos de gen��tipos puderam ser formados. Peso de massa seca de sementes, redu����o de reserva de sementes e peso de massa seca de pl��ntulas est��o positivamente correlacionados entre si, enquanto a redu����o de reserva das sementes e a efici��ncia na convers��o dessas reservas em pl��ntulas est��o negativamente correlacionadas. De acordo com os resultados obtidos, verificou-se que todas as caracter��sticas apresentaram n��veis diferenciados de variabilidade gen��tica e os crit��rios de sele����o utilizados mostraram-se eficientes para o melhoramento, no qual o ��ndice de sele����o combinada �� o crit��rio de sele����o que apresentou os melhores resultados em termos de ganhos, sendo indicado como crit��rio mais apropriado para o melhoramento gen��tico da popula����o estudada. Nos estudos de correla����es, em 70% dos casos, a correla����o fenot��pica foi superior �� genot��pica, mostrando maior influ��ncia dos fatores ambientais em rela����o aos genot��picos e condi����es prop��cias ao melhoramento dos diferentes caracteres. No estudo de diverg��ncia gen��tica, observou-se que pelo agrupamento de gen��tipos, pela t��cnica de Tocher, indicou que os gen��tipos foram distribu��dos em tr��s grupos.

Reconstitui����o das vaz��es naturais a jusante das pequenas centrais hidrel��tricas Rio Bonito e Su����a : Rio Santa Maria da Vit��ria (ES)

O conhecimento do regime natural de rios �� relevante, entre diversas raz��es, para avaliar o impacto antr��pico em bacias hidrogr��ficas. Desta forma, visa criar ferramentas que auxiliem no planejamento eficiente dos recursos h��dricos. O rio Santa Maria da Vit��ria (ES) �� crucial para o desenvolvimento socioecon��mico da Regi��o Metropolitana da Grande Vit��ria. Adicionalmente, diversos ecossistemas e comunidades agr��colas dependem deste rio. Sendo a gera����o de energia el��trica um dos usos mais significativos do rio Santa Maria da Vit��ria, atrav��s das Pequenas Centrais Hidrel��tricas Rio Bonito e Su����a, o objetivo principal desta pesquisa foi buscar uma melhor compreens��o da influ��ncia da opera����o destes empreendimentos na vaz��o natural deste rio. Al��m da an��lise das s��ries de vaz��es e de precipita����es medidas em Santa Leopoldina, comparando o per��odo anterior com o posterior �� implanta����o das PCH`s, realizou-se a reconstitui����o das vaz��es naturais m��dias di��rias e m��dias mensais a jusante das PCH`s, atrav��s da t��cnica de extens��o MOVE.I (tipo 1) e com o aux��lio da s��rie de vaz��o medida na esta����o fluviom��trica Marechal Floriano, localizada na bacia hidrogr��fica do rio Jucu. Os resultados obtidos, a partir dos dados monitorados na esta����o Santa Leopoldina, mostraram uma diminui����o nas vaz��es m��nimas com 7, 30 e 90 dias de dura����o para o per��odo p��s-regulado. Ainda, as vaz��es de perman��ncia Q50, Q90 e Q95 e a vaz��o m��dia de longa dura����o reconstitu��das s��o menores que as vaz��es observadas em Santa Leopoldina. Al��m da presen��a das PCH`s, as altera����es no uso e cobertura do solo da bacia podem ter influ��ncia sobre os resultados encontrados. As vaz��es em Santa Leopoldina, devido aos hor��rios em que s��o realizadas as medi����es, podem n��o representar de forma confi��vel as oscila����es a jusante das PCH`s. Consequentemente, faz-se necess��rio um monitoramento cont��nuo das vaz��es a jusante desses empreendimentos e recomenda-se que seja realizada uma nova reconstitui����o das vaz��es naturais do rio Santa Maria da Vit��ria, atrav��s do balan��o h��drico dos reservat��rios, utilizando os dados operativos das PCH`s.

Cen��rios de mudan��as clim��ticas para a Am��rica do Sul : impacto no zoneamento agroclim��tico da cana-de-a����car

O Painel Intergovernamental sobre Mudan��as Clim��ticas (IPCC) atrav��s do seu Quarto Relat��rio de Avalia����o das Mudan��as Clim��ticas Globais (IPCC-AR4), publicado em 2007, atribui as emiss��es de gases de efeito estufa como a principal causa do aumento m��dio das temperaturas e alerta para uma eleva����o entre 1,8 ��C e 6,4 ��C at�� 2100, podendo modificar assim a aptid��o clim��tica para as culturas agr��colas em diversas regi��es do planeta. Diante disso, existe a necessidade de substitui����o dos combust��veis f��sseis por fontes renov��veis e limpas de energia, como o etanol. A cana-de-a����car apresenta-se, portanto, como uma cultura estrat��gica na produ����o do etanol. O presente trabalho teve como objetivos: 1) avaliar o desempenho dos Modelos Clim��ticos Globais (MCGs) do IPCC-AR4 na simula����o de dados clim��ticos de temperatura do ar e precipita����o pluviom��trica para o per��odo anual e mensal; 2) elaborar o zoneamento agroclim��tico da cana-de-a����car para a Am��rica do Sul considerando o clima refer��ncia e o futuro para as d��cadas de 2020, 2050 e 2080 em fun����o do cen��rio de emiss��o A1B considerado pessimista e que usa um equil��brio entre todas as fontes de energia. Para a avalia����o do desempenho dos MCGs, foram utilizados dados clim��ticos m��dios mensais observados de precipita����o e temperatura do ar provenientes do Climatic Research Unit (CRU) e dados simulados oriundos dos 22 MCGs do IPCC (cen��rio 20c3m) compreendidos entre o per��odo de 1961-1990, al��m do Multimodel (ensemble) ��� MM que �� a m��dia da combina����o dos dados de todos os modelos. O desempenho dos MCGs foi avaliado pelos ��ndices estat��sticos: desvio padr��o, correla����o, raiz quadrada da m��dia do quadrado das diferen��as centralizadas e o ���bias��� dos dados simulados com os observados, que foram representados no diagrama de Taylor. Para a etapa da elabora����o do zoneamento agroclim��tico procedeu-se o c��lculo dos balan��os h��dricos (refer��ncia e futuros) da cultura, pelo m��todo de Thornthwaite & Mather (1955). Para o cen��rio refer��ncia, utilizaram-se dados das m��dias mensais da precipita����o e temperatura provenientes do CRU, enquanto que para as proje����es futuras, dados provenientes das anomalias do Multimodel (ensemble) ��� MM para as d��cadas de 2020, 2050 e 2080, que foram ajustados, obtendo-se assim as proje����es futuras para cada per��odo analisado. Baseado nos mapas tem��ticos reclassificados de defici��ncia h��drica anual, temperatura m��dia anual, excedente h��drico anual e no ��ndice de satisfa����o das necessidades de ��gua (ISNA), realizou-se uma sobreposi����o dessas informa����es obtendo assim, os mapas finais do zoneamento agroclim��tico da cana-de-a����car. Posteriormente ao zoneamento, realizou-se a an��lise das transi����es (ganhos, perdas e persist��ncias) entre as classes de aptid��o clim��tica da cultura. Os resultados mostram que o Multimodel (ensemble) ��� MM para o per��odo mensal apresenta o melhor desempenho entre os modelos analisados. As ��reas inaptas correspondem a maior parte da Am��rica do Sul e uma expressiva transi����o entre as classes de aptid��o clim��tica da cultura.

Eventos de vida em mulheres com diagn��stico de c��ncer de mama

Introdu����o: A rela����o entre eventos de vida e o surgimento e progress��o do c��ncer de mama tem sido investigada por alguns estudos. Objetivo: Identificar os eventos de vida ocorridos em mulheres com diagn��stico de c��ncer de mama, examinar o tempo transcorrido entre o evento e o diagn��stico, examinar a associa����o entre a sobrecarga ocasionada pelo evento no momento da ocorr��ncia e ap��s o diagn��stico e examinar a associa����o entre met��stase e eventos de vida p��s diagn��stico de c��ncer de mama. Metodologia: Estudo transversal realizado no Hospital Santa Rita de C��ssia, Vit��ria ��� ES. Comp��e-se a amostra por 300 mulheres. Coletaram-se os dados no per��odo de setembro a dezembro de 2014. Utilizou-se o instrumento Life Events Units ��� LEU/VAS que se baseia na Escala de Avalia����o de Reajustamento Social de Holmes e Rahe, que no Brasil foi adaptada por Vasconcellos. Utilizou-se o Pacote Estat��stico para Ci��ncias Sociais (SPSS), vers��o 20.0, para calcular a frequ��ncia, m��dia, mediana, desvio padr��o e aplicar os testes n��o param��trico de Wilcoxon e qui-quadrado. Resultados: A m��dia de idade foi de 53 anos. Predominou-se mulheres de ra��a/cor n��o branca (65%), com menos de 8 anos de estudo (64%) e casadas (54%). Identificou-se que a maioria da amostra relatou pelo menos um evento de vida (99,3%). O principal evento de vida relatado foi morte de algu��m na fam��lia. As medianas do tempo transcorrido entre os eventos de vida mais relatados e o diagn��stico de c��ncer de mama variaram de 5 a 15 anos. Observou-se diferen��a significante (p< 0,05) entre a sobrecarga ocasionada pelos eventos de vida nos dois momentos examinados. Em rela����o �� met��stase, 31,7% das que tiveram recidivas relataram um ou mais eventos de vida (p= 0,001). Quando considerado o tempo transcorrido entre o diagn��stico e o surgimento da met��stase nas 46 mulheres, observou-se uma mediana de 18,0 meses. Conclus��o: Os resultados deste estudo s��o potencialmente importantes, pois d��o suporte a uma poss��vel intera����o entre eventos de vida p��s diagn��stico de c��ncer de mama e met��stase. Estudos futuros s��o necess��rios para melhor compreens��o desta rela����o.

Intraocular pressure measurement in sheep using an applanation tonometer

The purpose of this study was to evaluate and establish the mean values of IOP in healthy adult sheep using an applanation tonometer. Information on age, sex, and breed was obtained for all animals included in this study. Twenty five healthy sheep (Ovis aries), of the same breed (Texel), male or female, with three years of age, received an ophthalmic examination in both eyes, including pupillary reflexes, Schirmer tear test, slit lamp biomicroscopy, and fluorescein staining. For all ophthalmic testing, animals were gently physically restrained, with no pressure in the jugular area and the eyelids were carefully open. IOP was measured by applanation tonometry (Tonopen XL). The same examiner performed the tonometry; measurements were taken three times for each eye, and their average was recorded as the IOP of the animal. Statistical analysis was performed using paired t-test and values of P < 0.05 were considered significant. The mean intraocular pressure in the whole group of 50 eyes was of 16.36 +/- 2.19 mm Hg. The mean (SD) IOP in the right eye was of 15.96 +/- 2.02 mm Hg, while the mean (SD) IOP in the left eye was of 16.76 +/- 2.32 mm Hg. Significant differences in IOP were not found between right and left eyes. The applanation tonometer was adequate for measuring the intraocular pressure in sheep. Reference data will assist in diagnosing testing for ophthalmic disease in sheep, as�� well as promote further studies in this area.

Comparative analysis between statistical and artificial intelligence models in business failure prediction

A growing number of predicting corporate failure models has emerged since 60s. Economic and social consequences of business failure can be dramatic, thus it is not surprise that the issue has been of growing interest in academic research as well as in business context. The main purpose of this study is to compare the predictive ability of five developed models based on three statistical techniques (Discriminant Analysis, Logit and Probit) and two models based on Artificial Intelligence (Neural Networks and Rough Sets). The five models were employed to a dataset of 420 non-bankrupt firms and 125 bankrupt firms belonging to the textile and clothing industry, over the period 2003���09. Results show that all the models performed well, with an overall correct classification level higher than 90%, and a type II error always less than 2%. The type I error increases as we move away from the year prior to failure. Our models contribute to the discussion of corporate financial distress causes. Moreover it can be used to assist decisions of creditors, investors and auditors. Additionally, this research can be of great contribution to devisers of national economic policies that aim to reduce industrial unemployment.