Clustering and selecting categorical features

In data clustering, the problem of selecting the subset of most relevant features from the data has been an active research topic. Feature selection for clustering is a challenging task due to the absence of class labels for guiding the search for relevant features. Most methods proposed for this goal are focused on numerical data. In this work, we propose an approach for clustering and selecting categorical features simultaneously. We assume that the data originate from a finite mixture of multinomial distributions and implement an integrated expectation-maximization (EM) algorithm that estimates all the parameters of the model and selects the subset of relevant features simultaneously. The results obtained on synthetic data illustrate the performance of the proposed approach. An application to real data, referred to official statistics, shows its usefulness.

Distribuições estatísticas, testes de ajuste e métodos para estimação de parâmetros na área da fiabilidade

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia Mecânica

Laboratory acquired infection by the virus SP H 114202 (Arenavirus: Arenaviridae): clinical and laboratory findings

Here in is described the clinical and laboratorial findings of a laboratory-acquired infection caused by the virus SP H 114202 (Arenavirus, family Arenaviridae) a recently discovered agent responsible for a viral hemorrhagic fever. The patient was sick for 13 days. The disease had an abrupt onset characterized by high fever (39ºC.), headache, chills and myalgias for 8 days. In addition, on the 3rd day, the patient developed nauseas and vomiting, and in the 10th, epigastralgia, diarrheia and gengivorrhagia. Leucopenia was seen within the 1 st week of onset, with counts as low as 2,500 white cells per mm³. Counts performed after the 23th day of the onset were within normal limits. With the exception of moderate lymphocitosis, no changes were observed in differential counts. An increase in the liter of antibodies by complement fixation, neutralization and ELISA (IgM) was detected. Suckling mice and baby hamsters were inoculated intracerebrally with 0.02 ml of blood samples collected in the 2nd and 7th days of disease. Attempts to isolate the virus were also made in Vero cells. No virus was isolated. This virus was isolated before in a single occasion in São Paulo State, in 1990, from the blood of a patient with hemorrhagic fever with a fatal outcome. The manipulation of the virus under study, must be done carefully, since the transmission can occur through aerosols.

Organised turbulence over mobile and immobile hydraulically rough boundaries

33rd IAHR Congress: Water Engineering for a Sustainable Environment

Aseptic meningitis in a large MMR vaccine campaign (590,609 people) in Curitiba, Paraná, Brazil, 1998

The aseptic meningitis after Measles-Mumps-Rubella vaccine (MMR) is a well recognized complication, and different incidences have been observed in several studies. We retrospectively analyzed forty cases of aseptic meningitis, during a large public immunization campaign (1998) in Curitiba, Southern Brazil (590,609 people), admitted in our Service. The vaccine utilized was Leningrad-3-Zagreb mumps strain, Edmonston-Zagreb measles strain, and RA 27#3 rubella strain. In all county, a total number of 87 cases were reported, resulting in a incidence of 1.7 cases per 10,000 given doses . The mean age was 23.7 ± 12.8 years. The female:male ratio was 1.35:1. Severe headache with meningismus (92.5%), fever (87.5%), nausea/vomiting (82.5%) were the most common clinical findings. Three cases (7.5%) developed mild mumps. All patients underwent cerebrospinal fluid (CSF) tap with the following findings: mononuclear pleocytosis from 100 to 500 cells/mm³ in 17 cases (42.5%; 257.5 ± 260.6 cells/mm³); increased protein 28 cases (67.5%; 92.1 ± 76.9 mg/dL); glucose was normal in all cases (56.8 ± 11.2 mg/dL) except in 4 (10%) cases, which presented less than 44 mg/dL. All serological tests (latex to bacterial meningitis, Cryptococcus, cysticercosis, VDRL) and bacteriological cultures were negative. Virus identification were also negative in 8 samples. None of the patients had neurological deficits or related symptoms after one year of onset. We believe the benefit of vaccination clearly outweights the incidence of benign vaccine-associated meningitis.

Estudo da resistência de juntas adesivas em L entre alumínio e material compósito

As ligações adesivas são frequentemente utilizadas na fabricação de estruturas complexas que não poderiam ou não seriam tão fáceis de ser fabricadas numa só peça, a fim de proporcionar uma união estrutural que, teoricamente, deve ser pelo menos tão resistente como o material de base. As juntas adesivas têm vindo a substituir métodos como a soldadura, e ligações parafusadas e rebitadas, devido à facilidade de fabricação, menor custo, facilidade em unir materiais diferentes, melhor resistência, entre outras características. Os materiais compósitos reforçados com fibra de carbono são amplamente utilizados em muitas indústrias, tais como de construção de barcos, automóvel e aeronáutica, sendo usados em estruturas que requerem elevada resistência e rigidez específicas, o que reduz o peso dos componentes, mantendo a resistência e rigidez necessárias para suportar as diversas cargas aplicadas. Embora estes métodos de fabricação reduzam ao máximo as ligações através de técnicas de fabrico avançadas, estas ainda são necessárias devido ao tamanho dos componentes, limitações de projecto tecnológicas e logísticas. Em muitas estruturas, a combinação de compósitos com metais tais como alumínio ou titânio traz vantagens de projecto. Este trabalho tem como objectivo estudar, experimentalmente e por modelos de dano coesivo (MDC), juntas adesivas em L entre componentes de alumínio e compósito de carbono epóxido quando solicitados a forças de arrancamento, considerando diferentes configurações de junta e adesivos de ductilidade distinta. Os parâmetros geométricos abordados são a espessura do aderente de alumínio (tP2) e comprimento de sobreposição (LO). A análise numérica permitiu o estudo da distribuição das tensões, evolução do dano, resistência e modos de rotura. Os testes experimentais validam os resultados numéricos e fornecem mecanismos de projecto para juntas em L. Foi mostrado que a geometria do aderente em L (alumínio) e o tipo de adesivo têm uma influência directa na resistência de junta.

Monte Carlo simulations for dosimetric verification in photon and electron beam radiotherapy

Dissertação para obtenção do Grau de Doutor em Engenharia Biomédica

Familial Amyloidotic Polineuropathy and Systemic Lupus

Familial amyloidotic polineuropathy is a genetic disorder, leading to systemic amyloid deposits, manifested as sensory-motor and autonomic neuropathy. In the Portuguese classical form, the disease is evident at a young age, and causes death if no specific treatment is received. Variability in penetrance, age of onset and clinical course has been published; environmental and genetic factors are believed to contribute to this variability. The authors report a case of a 51-year-old white female, with a medical history of acquired angioedema, late-onset familial amyloidotic polineuropathy and systemic lupus erythemathosus. The authors consider that these associated diseases could modulate their expression.

Neuromyelitis Optica in Portugal (NEMIPORT) - A Multicentre Study

BACKGROUND: Neuromyelitis Optica (NMO) is an inflammatory demyelinating disease of the CNS. There have been few epidemiologic studies on NMO, none in Portugal. OBJECTIVE: To analyze the clinical, biological and MRI characteristics from a cohort of Portuguese patients who fulfilled the Wingerchuk 2006 NMO/NMOSD criteria. To identify and characterize those who had concomitant autoimmune disease or circulating autoantibodies. METHODS: We performed an observational, retrospective, multicenter study in 5 Hospital Centers in Portugal. RESULTS: Sixty-seven patients fulfilled the inclusion criteria. They were mainly Caucasian, 55 female. Median age at onset was 32.0 years and mean follow-up 7.4±6.0 years. Twenty-one patients were definite NMO and optic neuritis (ON) the most frequent initial presentation. Forty-six were classified as NMO spectrum disorders. The main subtypes were recurrent ON and single longitudinally extensive transverse myelitis. Twenty-four patients had positive AQP4-IgG. Twenty-three had other circulating autoantibodies. Fifteen out of 67 patients had concomitant autoimmune disease. There was a significant correlation between the presence of autoimmune disease and the positivity for AQP4-IgG. Five patients died, all definite NMO. CONCLUSION: This is the first study about this rare disease in Portugal. Demographic features were similar to other studies. The existence of concomitant autoimmune disease was significantly associated with seropositivity for AQP4-IgG.

PROP1 Gene Analysis in Portuguese Patients with Combined Pituitary Hormone Deficiency

OBJECTIVE: Mutations of the PROP1 gene lead to combined pituitary hormone deficiency (CPHD), which is characterized by a deficiency of GH, TSH, LH/FSH, PRL and, less frequently, ACTH. This study was undertaken to investigate the molecular defect in a cohort of patients with CPHD. DESIGN, PATIENTS AND MEASUREMENTS: A multicentric study involving 46 cases of CPHD (17 familial cases belonging to seven kindreds and 29 sporadic cases) selected on the basis of clinical and hormonal evidence of GH deficiency, central hypothyroidism and hypogonadotrophic hypogonadism, in the absence of an identified cause of hypopituitarism. Mutations of PROP1 were investigated by DNA sequencing. Clinical, hormonal and neuroradiological data were collected at each centre. RESULTS: PROP1 mutations were identified in all familial cases: five kindreds presented a c. 301-302delAG mutation, one kindred presented a c. 358C --> T (R120C) mutation and one presented a previously unreported initiation codon mutation, c. 2T --> C. Of the 29 sporadic cases, only two (6.9%) presented PROP1 germline mutations (c. 301-302delAG, in both). Phenotypic variability was observed among patients with the same mutations, particularly the presence and age of onset of hypocortisolism, the levels of PRL and the results of pituitary imaging. One patient presented a sellar mass that persisted into adulthood. CONCLUSIONS: This is the first report of a mutation in the initiation codon of the PROP1 gene and this further expands the spectrum of known mutations responsible for CPHD. The low mutation frequency observed in sporadic cases may be due to the involvement of other unidentified acquired or genetic causes.

Pregnant crack addicts in a psychiatric unit

OBJECTIVE: In this study we aim to characterize a sample of 85 pregnant crack addicts admitted for detoxification in a psychiatric inpatient unit. METHOD: Cross-sectional study. Sociodemographic, clinical, obstetric and lifestyle information were evaluated. RESULTS: Age of onset for crack use varied from 11 to 35 years (median = 21). Approximately 25% of the patients smoked more than 20 crack rocks in a typical day of use (median = 10; min-max = 1-100). Tobacco (89.4%), alcohol (63.5%) and marijuana (51.8%) were the drugs other than crack most currently used. Robbery was reported by 32 patients (41.2%), imprisonment experience by 21 (24.7%), trade of sex for money/drugs by 38 (44.7%), home desertion by 33 (38.8%); 15.3% were positive for HIV, 5.9% for HCV, 1.2% for HBV and 8.2% for syphilis. After discharge from the psychiatric unit, only 25% of the sample followed the proposed treatment in the chemical dependency outpatient service. CONCLUSION: Greater risky behaviors for STD, as well as high rates of maternal HIV and Syphilis were found. Moreover, the high rates of concurrent use of other drugs and involvement in illegal activities contribute to show their chaotic lifestyles. Prevention and intervention programs need to be developed to address the multifactorial nature of this problem.

Non-classical error bounds in the central limit theorem

The classical central limit theorem states the uniform convergence of the distribution functions of the standardized sums of independent and identically distributed square integrable real-valued random variables to the standard normal distribution function. While first versions of the central limit theorem are already due to Moivre (1730) and Laplace (1812), a systematic study of this topic started at the beginning of the last century with the fundamental work of Lyapunov (1900, 1901). Meanwhile, extensions of the central limit theorem are available for a multitude of settings. This includes, e.g., Banach space valued random variables as well as substantial relaxations of the assumptions of independence and identical distributions. Furthermore, explicit error bounds are established and asymptotic expansions are employed to obtain better approximations. Classical error estimates like the famous bound of Berry and Esseen are stated in terms of absolute moments of the random summands and therefore do not reflect a potential closeness of the distributions of the single random summands to a normal distribution. Non-classical approaches take this issue into account by providing error estimates based on, e.g., pseudomoments. The latter field of investigation was initiated by work of Zolotarev in the 1960's and is still in its infancy compared to the development of the classical theory. For example, non-classical error bounds for asymptotic expansions seem not to be available up to now ...

Análise estatística da distribuição de Poisson

The general properties of POISSON distributions and their relations to the binomial distribuitions are discussed. Two methods of statistical analysis are dealt with in detail: X2-test. In order to carry out the X2-test, the mean frequency and the theoretical frequencies for all classes are calculated. Than the observed and the calculated frequencies are compared, using the well nown formula: f(obs) - f(esp) 2; i(esp). When the expected frequencies are small, one must not forget that the value of X2 may only be calculated, if the expected frequencies are biger than 5. If smaller values should occur, the frequencies of neighboroughing classes must ge pooled. As a second test reintroduced by BRIEGER, consists in comparing the observed and expected error standard of the series. The observed error is calculated by the general formula: δ + Σ f . VK n-1 where n represents the number of cases. The theoretical error of a POISSON series with mean frequency m is always ± Vm. These two values may be compared either by dividing the observed by the theoretical error and using BRIEGER's tables for # or by dividing the respective variances and using SNEDECOR's tables for F. The degree of freedom for the observed error is one less the number of cases studied, and that of the theoretical error is always infinite. In carrying out these tests, one important point must never be overlloked. The values for the first class, even if no concrete cases of the type were observed, must always be zero, an dthe value of the subsequent classes must be 1, 2, 3, etc.. This is easily seen in some of the classical experiments. For instance in BORKEWITZ example of accidents in Prussian armee corps, the classes are: no, one, two, etc., accidents. When counting the frequency of bacteria, these values are: no, one, two, etc., bacteria or cultures of bacteria. Ins studies of plant diseases equally the frequencies are : no, one, two, etc., plants deseased. Howewer more complicated cases may occur. For instance, when analising the degree of polyembriony, frequently the case of "no polyembryony" corresponds to the occurrence of one embryo per each seed. Thus the classes are not: no, one, etc., embryo per seed, but they are: no additional embryo, one additional embryo, etc., per seed with at least one embryo. Another interestin case was found by BRIEGER in genetic studies on the number os rows in maize. Here the minimum number is of course not: no rows, but: no additional beyond eight rows. The next class is not: nine rows, but: 10 rows, since the row number varies always in pairs of rows. Thus the value of successive classes are: no additional pair of rows beyond 8, one additional pair (or 10 rows), two additional pairs (or 12 rows) etc.. The application of the methods is finally shown on the hand of three examples : the number of seeds per fruit in the oranges M Natal" and "Coco" and in "Calamondin". As shown in the text and the tables, the agreement with a POISSON series is very satisfactory in the first two cases. In the third case BRIEGER's error test indicated a significant reduction of variability, and the X2 test showed that there were two many fruits with 4 or 5 seeds and too few with more or with less seeds. Howewer the fact that no fruit was found without seed, may be taken to indicate that in Calamondin fruits are not fully parthenocarpic and may develop only with one seed at the least. Thus a new analysis was carried out, on another class basis. As value for the first class the following value was accepted: no additional seed beyond the indispensable minimum number of one seed, and for the later classes the values were: one, two, etc., additional seeds. Using this new basis for all calculations, a complete agreement of the observed and expected frequencies, of the correspondig POISSON series was obtained, thus proving that our hypothesis of the impossibility of obtaining fruits without any seed was correct for Calamondin while the other two oranges were completely parthenocarpic and fruits without seeds did occur.

Procedurally fair and stable matching

We motivate procedural fairness for matching mechanisms and study two procedurally fair and stable mechanisms: employment by lotto (Aldershof et al., 1999) and the random order mechanism (Roth and Vande Vate, 1990, Ma, 1996). For both mechanisms we give various examples of probability distributions on the set of stable matchings and discuss properties that differentiate employment by lotto and the random order mechanism. Finally, we consider an adjustment of the random order mechanism, the equitable random order mechanism, that combines aspects of procedural and "endstate'' fairness. Aldershof et al. (1999) and Ma (1996) that exist on the probability distribution induced by both mechanisms. Finally, we consider an adjustment of the random order mechanism, the equitable random order mechanism.

Welfare comparisons: sequential procedures for heterogenous populations

Some analysts use sequential dominance criteria, and others use equivalence scales in combination with non-sequential dominance tests, to make welfare comparisons of oint distributions of income and needs. In this paper we present a new sequential procedure hich copes with situations in which sequential dominance fails. We also demonstrate that there commendations deriving from the sequential approach are valid for distributions of equivalent income whatever equivalence scale the analyst might adopt. Thus the paper marries together the sequential and equivalizing approaches, seen as alternatives in much previous literature. All results are specified in forms which allow for demographic differences in the populations being compared.