Dans une turbine hydraulique, la rotation des aubes dans l’eau crée une zone de basse pression, amenant l’eau à passer de l’état liquide à l’état gazeux. Ce phénomène de changement de phase est appelé cavitation et est similaire à l’ébullition. Lorsque les cavités de vapeur formées implosent près des parois, il en résulte une érosion sévère des matériaux, accélérant de façon importante la dégradation de la turbine. Un système de détection de l’érosion de cavitation à l’aide de mesures vibratoires, employable sur les turbines en opération, a donc été installé sur quatre groupes turbine-alternateur d’une centrale et permet d’estimer précisément le taux d’érosion en kg/ 10 000 h. Le présent projet vise à répondre à deux objectifs principaux. Premièrement, étudier le comportement de la cavitation sur un groupe turbine-alternateur cible et construire un modèle statistique, dans le but de prédire la variable cavitation en fonction des variables opératoires (tels l’ouverture de vannage, le débit, les niveaux amont et aval, etc.). Deuxièmement, élaborer une méthodologie permettant la reproductibilité de l’étude à d’autres sites. Une étude rétrospective sera effectuée et on se concentrera sur les données disponibles depuis la mise à jour du système en 2010. Des résultats préliminaires ont mis en évidence l’hétérogénéité du comportement de cavitation ainsi que des changements entre la relation entre la cavitation et diverses variables opératoires. Nous nous proposons de développer un modèle probabiliste adapté, en utilisant notamment le regroupement hiérarchique et des modèles de régression linéaire multiple. / Cavitation erosion which results from repeated collapse of transient vapor cavities on solid surfaces is a constant problematic in hydraulic turbine runners and continues to enforce costly repair and loss of revenues. A vibratory detection system of cavitation erosion was installed 10 years ago for continuous monitoring of 4 hydropower units. A new hardware version of the system was developed and installed in 2010. This new system configuration is more reliable and allows more accurate evaluation of the cavitation erosion of the runners in kg/10 000 h. The first objective of this study is to investigate cavitation behavior upon one generating unit and to build a statistical model which will allow prediction of instant cavitation related to operating variables, such as gate opening, water flow, headwater level, tailwater levels, etc. The second objective is to develop a methodology for the reproducibility of the studies to other sites. A retrospective study will be conducted and we will mainly focus on data available since the system update in 2010. The preliminary analysis enhanced the complexity of the phenomenon. Indeed, changes in the relationship between cavitation and various operating variables were observed and could be due to a seasonal behavior or different operating conditions. Using hierarchical clustering and regression models, we formalize this heterogeneity by developing a model which includes operating variables such as active power, tailwater level and gate opening.

Master of Science / Department of Plant Pathology / William Bockus / Barley yellow dwarf (BYD) is one of the most important wheat diseases in the state of Kansas. Despite the development of cultivars with improved levels of resistance to BYD, little is known about the impact that this resistance has on yield loss from the disease. The intent of this research was to estimate yield loss in winter wheat cultivars in Kansas due to BYD and quantify the reduction in losses associated with resistant cultivars. During seven years, BYD disease incidence was visually assessed on numerous winter wheat cultivars in replicated field nurseries. Cultivars were planted about three weeks early to promote disease. When grain yields were regressed against BYD incidence scores, negative linear relationships significantly fit the data for each year and for the combined dataset covering all seven years. The models showed that, depending upon the year, 19-48% (average 33%) of the yields was explained by BYD incidence. For the combined dataset, 29% of the relative yield was explained by BYD incidence. The models predicted that cultivars showing high disease incidence had 25-86% (average 49%) less yield than a hypothetical cultivar that showed zero incidence. Using the models, the moderate level of resistance in the cultivar Everest was calculated to reduce yield loss from BYD by about 73%. Therefore, utilizing visual BYD symptom evaluations in Kansas, coupled with grain yields, is useful to estimate yield loss from the disease. Furthermore, linear models that incorporate those parameters can be used to calculate the impact of improving cultivar resistance to BYD on yield losses.

The purpose of this thesis is to investigate further whether oil price has an effect on house price inflation and additionally if it has a link to house price turning points. The methodology is grounded on the previous research paper made by Breitenfellner et al. (2015). The results are based on quarterly data from the countries; Finland, Denmark, Norway and Sweden through the time span of 1990-2018. A linear fixed regression model was performed including the explanatory variables of monetary policy and credit developments, macroeconomic fundamentals, housing market variable and demographic variables. Secondly, a logit model was used to identify a relationship between oil price and house price turning points. The model used misalignment made from GDP per capita and real interest rate. The empirical analysis confirms that there is a positive relationship between oil prices and house price inflation. This evidence contradicts a major share of previous research papers (see Bernanke, 2010; Kaufmann et al., 2011). However, there are also some previous papers (see Yiqi, (2017); Antonakakis et al., 2016) and theoretical linkages in line with a positive correlation. Concerning, the oil price and house price inflation no empirical significance was found regarding their relationship. For future research, one could include regional aspects for the purpose of controlling for geographical differences.

During the last few decades, the Swedish population has shown an increased interest in investment fund savings, and more than 60% of Swedish citizens are saving through funds today. In addition to this, awareness on climate change and related risks has increased, which has contributed to a greater focus on corporate sustainability among Swedish companies. As a result of these trends, there has been an increase in fund companies that are basing their investments on certain ethical restrictions, in order for private investors to save ethically. The purpose of this study was to examine how ethical restrictions affect the financial performance of Swedish funds with regards to risk-adjusted return. Previous studies have focused on comparing ethical and traditional funds, but this study chose not to make any difference between the two types of funds. The study's relevance is based on how it can make private investors aware of which ethical restrictions that have a negative impact on the risk-adjusted return, and how much each restriction decreases the return. A total of 101 Swedish funds were analyzed. Information about each fund's performance measures were retrieved from Morningstar, and is based on three years development. Information about each fund's ethical restrictions were retrieved from Hållbarhetsprofilen and their information pamphlets. With data about the performance measures Sharpe ratio, Alpha, and Treynor ratio, three statistical models were defined and analyzed with multiple linear regression analysis. Each model's reliability was assessed with residual analysis; the models were adjusted and improved if necessary. Hypotheses were evaluated with significance testing to answer the research questions. The results indicate that exclusion of tobacco and gambling companies affect the risk-adjusted return of Swedish funds negatively, while exclusion of alcohol companies affect the risk-adjusted return positively. This implies that private investors should save their money in Swedish funds that exclude alcohol companies in order to avoid lower risk-adjusted return. / Fondsparandet i Sverige har under de senaste decennierna ökat kraftigt och idag sparar över 60% av befolkningen i fonder. Utöver detta har bland annat miljömedvetenheten hos befolkningen ökat, vilket har bidragit till ett ökat engagemang i hållbarhetsfrågor hos företag. I kölvattnet av dessa trender har allt fler svenska fondbolag valt att införa etiska restriktioner på deras investeringar, så att fondsparare på ett enkelt sätt ska kunna investera hållbart. Syftet med denna studie var att undersöka hur etiska restriktioner påverkar svenska fonders prestation med avseende på riskjusterad avkastning. Där tidigare forskning har fokuserat på att jämföra etiska och traditionella fonder, valde denna studie istället att inte göra någon skillnad mellan de två fondtyperna. Studiens relevans baseras på hur studiens slutsatser kan hjälpa småsparare att bli medvetna om vilka etiska restriktioner som hämmar avkastningen och vad varje restriktion kostar investerare i potentiellt förlorad avkastning. Totalt har 101 svenska fonder undersökts. Information som gäller fondernas prestationsmått har hämtats från Morningstar, och baseras på tre års utveckling. Information om fondernas etiska restriktioner har hämtats från Hållbarhetsprofilen och fondernas informationsbroschyrer. Med insamlad data om prestationsmåtten Sharpekvoten, Alfa, och Treynorkvoten, har tre modeller definierats som sedan analyserats med multipel linjär regressionsanalys. Modellernas tillförlitlighet har sedan utvärderats med residualanalys; modellerna förbättrades om det fanns belägg för det. Utifrån signifikanstest har hypoteser utvärderats för att besvara studiens frågeställningar. Studiens resultat pekar på att exkludering av företag inom tobak- och spelbranschen påverkar svenska fonders riskjusterade avkastning negativt, medan exkludering av företag inom alkoholbranschen påverkar den positivt. Detta betyder alltså att småsparare bör investera i svenska fonder som väljer bort företag inom alkoholbranschen för att undvika låg avkastning.

Dans un contexte de mobilité durable, le vélo en libre-service (VLS) représente un élément important dans les politiques de mobilité urbaine pour favoriser les modes doux dans le centre urbain. Son développement rapide dans le monde nécessite des recherches pour mieux comprendre le succès de ce mode de déplacement. Notre contribution à travers cette recherche est d’analyser les éléments socio-économiques importants ainsi que les caractéristiques du système VLS qui influent son usage pour construire des modèles de prédiction de la demande en VLS. Partant de l’hypothèse que les éléments socio-économiques autour des stations de VLS constituent des facteurs qualitativement et quantitativement explicatifs pour l’usage du VLS, notre démarche de recherche repose d’abord sur une analyse statistique, temporelle et spatiale des éléments explicatifs du système VLS pour ensuite les quantifier. Ces éléments sont utilisés pour modéliser la génération de la demande journalière en VLS par la méthode de régression linéaire multiple. Les modèles estimant les flux entrants et sortants journaliers en VLS ont été développés. Ils sont basés sur l’environnement socio-économique autour des stations et des variables d’offre de VLS. Ces modèles sont utiles pour comprendre le fonctionnement du système VLS, l’améliorer et estimer la demande des nouvelles stations dans une perspective d’élargissement d’un système VLS existant. Les modèles développés peuvent être également utilisés pour la localisation et le dimensionnement des stations d’un nouveau système VLS. La démarche de détermination, de quantification des variables explicatives et de modélisation forment un cadre de travail pour modéliser la demande d’autres modes de déplacement partagés. / In a context of sustainable transport, the bike sharing is an important factor in the policies to promote soft modes of transport in the urban center. Its rapid development in the world requires a need to deepen the usage of this mode of travel. Our contribution through this research is to analyze the important built-environment factors and the characteristics of bike sharing system that influence the use of bike sharing in order to build predictive models of demand for bike sharing. Assuming that socio-economic elements around the bike sharing stations are qualitatively and quantitatively explanatory for the use of bike sharing, our approach are based primarily on statistical analysis of temporal and spatial elements explaining bike sharing usage in order to determine and quantify the important built-environment variables. These variables are then used to model the generation of the daily demand of bike sharing using multiple linear regression method. The models estimating inflows and outflows of bike sharing using socioeconomic variables determined in a buffer area of each station are built. These models are useful for estimating the demand for new stations in an enlargement perspective of an existing bike sharing system or location and sizing of a new bike sharing system. The similar process of determination, quantification of the explanatory variables and modelling can be used to form a framework to predict the demand of other vehicle sharing systems.

As linhas de transmissão no Brasil são geralmente muito extensas, uma vez que os grandes centros de consumo de energia ficam distantes das usinas hidrelétricas. Por essa razão, a construção e manutenção de linhas de transmissão é de grande importância e, em uma grande porcentagem destas linhas, as estacas helicoidais são usadas como fundações. No entanto, a previsão da profundidade final de instalação destas estacas ainda é um grande desafio para os projetistas, fornecedores de estacas helicoidais e construtoras. A profundidade final destas fundações é controlada pelo torque de instalação; portanto, se o torque necessário para instalar uma estaca pudesse ser calculado com acurácia, com base em suas dimensões, e nos resultados de ensaios de investigação de solo in situ (SPT), a previsão de comprimentos de estaca para estimativas de custos, a definição de equipamentos adequados para instalação e a estimativa da quantidade de seções de estacas a serem transportadas para uma determinada linha de transmissão seriam mais acuradas. Além disso, a capacidade de carga de estacas helicoidais pode ser estimada usando os resultados do torque final de instalação. Sendo assim, para atender à necessidade de se determinar o torque de instalação de fundações helicoidais ainda na fase de projeto, um método simplificado foi desenvolvido e validado com os resultados de 752 estacas helicoidais multi-hélices instaladas em solos predominantemente arenosos, de uma linha de transmissão brasileira. O modelo desenvolvido baseia-se nos resultados de ensaios SPT e considera o efeito de instalação das estacas no solo penetrado. Nesta dissertação é apresentada uma descrição detalhada do método proposto e uma comparação entre os resultados medidos em campo e calculados pelo método. Os resultados da comparação mostram que o método proposto pode estimar com sucesso o torque de instalação de estacas helicoidais. / The transmission lines in Brazil are usually very extensive, since the centers of power consumption are often far from the most hydroelectric plants. For this reason, the construction and maintenance of transmission lines is frequent in this country, and in a large percentage of transmission lines, helical piles are used as guy wire anchors and foundations for transmission towers. However, the estimates of the final embedded depth of numerous helical piles to be used in several towers of the transmission lines is still a challenge for the designers, pile suppliers and contractors. The final depth of helical foundations is controlled by the installation torque; therefore, if the torque necessary to install a helical pile could be accurately calculated based on the pile dimensions and results of in-situ soil tests (SPT), the prediction of pile lengths for costs estimations, the definition of suitable equipment for pile installation, and the estimate of the quantity of helical piles sections to be transported for a particular transmission line would be more precise. Additionally, the uplift capacity of helical piles can be estimated using the results of final installation torque. In order to address the existed need of determining the installation torque of helical foundations during the design phase, a simplified method was developed and validated with the results of 753 multi-helix piles installed in predominantly sandy soils of a Brazilian transmission line. The model proposed is based on the results of SPT tests, and considers the installation effect of helical piles on the soil penetrated. This text presents a detailed description of such method and a comparison between measured and predicted results. The comparison shows that the method proposed can successfully estimate the installation torque of helical piles.

O inventário florestal é uma das principais ferramentas na gestão dos recursos florestais, uma vez que as informações geradas por ele são utilizadas ao longo de toda a cadeia produtiva do setor. Desta forma, erros nas estimativas volumétricas dos inventários florestais devem ser controlados. Inúmeras informações podem ser obtidas a partir de imagens orbitais ou aerotransportadas, uma vez que podem cobrir facilmente toda a área de interesse, e estão comumente disponíveis em empresas florestais ou ao usuário final. A utilização de preditores derivados das imagens pode trazer benefícios para as estimativas do inventário florestal. Desta forma, a aplicação de técnicas de regressão linear múltipla (RLM) ganhou espaço no setor devido a sua facilidade de aplicação. Porém, a RLM não leva em consideração a dependência espacial entre as unidades amostrais, sendo que a geoestatística pode ser utilizada para predizer a distribuição espacial do estoque de madeira (VTCC) para uma dada região. A modelagem geoestatística mais simples como a krigagem ordinária (KO), por considerar apenas a dependência espacial entre os pontos não amostrados, pode apresentar erros de predição nestes locais. Tais erros podem ser reduzidos com a aplicação de técnicas mais robustas como a Krigagem com Deriva Externa (KDE), pois esta agrega as informações obtidas das imagens com a distribuição espacial do volume. Buscando-se avaliar as vantagens da integração do Sensoriamento Remoto (SR) ao inventário florestal foram testados 4 tipos diferentes de imagens; as oriundas dos satélites LANDSAT8, RAPIDEYE e GEOEYE, e as provenientes de aeronaves (Imagens Aerotransportadas). Avaliou-se também diferentes tipos de estimativas para a predição volumétrica sendo estas RLM, KDE e KO. A melhor estimativa serviu de variável auxiliar para o estimador de regressão (ER), sendo os resultados comparados com a abordagem tradicional da amostragem aleatória simples (AAS). Os resultados demonstraram por meio da validação cruzada que as estimativas da KDE foram mais eficientes que as estimativas da KO e da RLM. Os melhores preditores (variáveis auxiliares) foram aqueles derivados do satélite LANDSAT8 e do satélite RAPIDEYE. Obteve-se como produto das estimativas de KDE e RLM mapas capazes de detectar áreas com mortalidade ou anomalias em meio a formação florestal. A utilização de uma estimativa de KDE utilizando imagens LANDSAT8 como medida auxiliar para o ER permitiu reduzir o erro amostral da AAS de 3,87% para 2,34%. Da maneira tradicional, tal redução de erro apenas seria possível com um aumento de mais 99 unidades amostrais. / Forest Inventory (FI) is one of the main tools for managing forest resources, once the information derived from FI is used along the sector production chain. When estimating volume, errors resulting from FI are common, therefore these errors must be controlled. Once orbital or airborne imaging data are easily acquired for an entire area, and are commonly available in forest companies or for the end user, much information can be obtained from these products. The use of predictor derived from images can be of significant benefits to forest inventory estimates. For that reason, the application of linear multiple regression (LMR) techniques have taken place in the forest sector, due to the facilities of its application. However, the LMR technique does not take the spatial dependence among sample units in consideration, the geostatistics utilized to predict the spatial distribution of the wood stock (VTCC) for a specific region. Simpler geostatistical modeling as the ordinary kriging (OK), just takes in consideration the spatial dependence among non-sampled points, because of that, prediction errors can be found. Such errors can be reduced when techniques that are more robust are applied, such as the kriging with external drift (KED) approach. This technique aggregates the information obtained from the images with the spatial distribution of the volume. In order to evaluate the advantages of Remote Sensing and Forest Inventory integration, we considered 4 different types of images, from the satellites LANSAT 8, RAPIDEYE, GEOEYE and from airborne images. When predicting volume, three different approaches were evaluated: LMR, EDK, OK. The best model among those evaluated, served as auxiliary variable for the regression estimator (RE). The result were then compared to the traditional approach, simple random sampling (SRS).This approach showed, through a cross-validation, that the KDE estimates were more efficiently than the OK and the LMR. The best predictor model (auxiliary variables) were derived from LADNSAT 8 and RAPIDEYE satellites. There is a significant advantage to using the KDE and LMR approaches, as it allows for a spatial representation of areas with mortality or anomalies, in a forest environment. The combination of KDE approach and LANDSAT 8 images as an auxiliary method for the RE, abled the decrease of the sampling error of SRS from 3.87% to 2.34%.The traditional approaches to conduct plantation inventories would allow for this error reduction, only if there were an increase of 99 more sampling units.

A distribuição gama generalizada (GG) possui, como casos particulares, distribuição Weibull, log-normal, gama, qui-quadrado, entre outras. Por essa razão, ela e considerada uma distribuição exvel no ajuste dos dados. A ideia de Cordeiro e Castro (2011) foi utilizada para o desenvolvimento de duas novas distribuições de probabilidade a partir da distribuição GG. Uma delas e denominada de Kumaraswamy gama generalizada (KumGG) e possui cinco parâmetros; a outra distribuição e uma modificação de um dos parmetros de forma da distribuição KumGG e foi denominada de distribuição Kumaraswamy gama generalizada estendida (KumGGE). Desenvolveu-se o modelo de regressão log-Kumaraswamy gama generalizada estendida. Alem disso, a ideia de Adamidis e Loukas (1998) para modicar distribuições foi utilizada para a distribuição GG; essa nova distribuição foi nomeada de gama generalizada geometrica (GGG). A vantagem desses novos modelos reside na capacidade de acomodar varias formas da função risco eles tambem se mostraram uteis na discriminação de modelos. Para cada um dos modelos foram calculados os momentos, função geradora de momentos, os desvios medios, a conabilidade e a função densidade de probabilidade da estatistica de ordem. Para a estimação dos parâmetros, foram utilizados os metodos de maxima verossimilhanca e bayesiano e, finalmente, para ilustrar a aplicação das novas distribuições foram analisados alguns conjuntos de dados reais. / The generalized gamma (GG) distribution has as particular cases the Weibull, log-normal, gamma and Chi-square distributions, among others. For this reason, it is considered a exible distribution for tting data. In this paper, the idea of Cordeiro and Castro (2011) is used to develop two new probability distributions based on the GG distribution. The rst is called the generalized gamma Kumaraswamy (KumGG) and has ve parameters, while the other involves a modication of one of the shape parameters of the KumGG distribution and is called the extended generalized gamma Kumaraswamy (KumGGE). Based in these, we develop the extended generalized log-Kumaraswamy regression model. Besides this, we employ the idea regarding modifying distributions of Adamidis and Loukas (1998) for the GG distribution, calling this new distribution the geometric generalized gamma (GGG). The advantage of these new models rests in their capacity to accommodate various risk function forms. They are also useful in model discrimination. We calculate the moments, moments generating function, mean deviations, reliability and probability density function of the order statistics. To estimate the parameters we use the maximum likelihood and Bayesian methods. Finally, to illustrate the application of the new distributions, we analyze some real data sets.

Este trabalho é um estudo ecológico descritivo sobre a leishmaniose tegumentar americana (LTA) na região do Vale do Ribeira paulista (composto por 23 municípios), que teve como objetivo correlacionar a incidência da doença com fatores determinantes ambientais e sociais. O trabalho se refere ao período de 1998 a 2006 o qual foi divido em três triênios, o primeiro de 1998 a 2000, o segundo de 2001 a 2003 e o terceiro de 2004 a 2006. Os coeficientes de incidência da LTA foram calculados para estes períodos (Coefmtr1, Coefmtr2 e Coefmtr3) segundo Szklo. Essa divisão de triênios se deu de acordo com as características gráficas dos coeficientes de incidência da LTA no período, que demonstraram haver num primeiro período uma fase endêmica da doença, no segundo período (2001 a 2003) a análise mostra picos de incidência que caracteriza um período epidêmico acentuado. No terceiro período alguns municípios mostram picos de incidência de proporções menores que os observados no período anterior, contudo ainda característicos de fase epidêmica. Os coeficientes de incidência são as variáveis dependentes dos três modelos de regressão realizados. As variáveis independentes foram obtidas de fontes de dados secundários e são de natureza sócio-demográfica, relacionadas as condições climáticas, de natureza político-geográficas e físico-biológicas. Essas variáveis foram separadas em qualitativas e quantitativas e analisadas respectivamente por testes não paramétricos de Mann Whitney e Spearman e selecionou-se para o processo de modelagem múltipla, as variáveis com p<0,20. Houve a necessidade de três modelos diferentes para verificar se os fatores determinantes que levaram as séries a se comportar de maneira diferenciada, seriam os mesmos ou não. O que se percebe é que há uma mudança no padrão do primeiro para os dois outros períodos. O trabalho aponta para fatores sócio-demográficos como determinantes na incidência de LTA, bem como a presença do vetor / This work is an ecological descriptive study on the American cutaneous leishmaniasis (ACL) in the Ribeira Valley region of Sao Paulo (composed of 23 municipalities), which aimed to correlate the incidence of the disease with environmental and social factors. The paper refers to the period 1998 to 2006 which was divided into three years, the first from 1998 to 2000, the second from 2001 to 2003 and the third from 2004 to 2006. The rates of incidence of ACL were calculated for these periods (Coefmtr1, Coefmtr2 and Coefmtr3) Szklo seconds. This division was made in accordance with the graphics characteristics of the coefficients of incidence of ACL in the period, which has shown a period in a first stage of the disease endemic in the second period (2001 to 2003) the peak incidence analysis shows that characterizes a epidemic period strong. In the second period some municipalities show peaks of incidence of proportions lower than those observed in the previous period, yet still features the epidemic stage. The rates of incidence are dependent variables of the three regression models performed. The independent variables were obtained from sources in secondary databases, and are: in nature and socio-demographic, related climatic conditions, a political-geographical and physical and biological. These variables were separated in quality and quantity respectively and analyzed by non-parametric tests for Spearman and Mann Whitney and selected for the process of regress multiple model, the variables with p <0.20. There was a need for three different models to see if the determining factors that led to the series to behave in a different way, would be the same or not. What we understand is that there is a change in the pattern of the first to the other two periods. The study points to socio-demographic factors as determinants in the incidence of ACL and the presence of vector

Introdução: A avaliação dos resultados de políticas públicas de suplementação alimentar em condições reais de sua operacionalização (efetividade) é um instrumento imprescindível para área de saúde pública. Objetivos: Avaliar a efetividade de programa governamental de suplementação alimentar no crescimento de crianças, segundo o estado nutricional ao ingressar. Métodos: Estudo de coorte com dados secundários de 25.433 crianças de baixa renda com idade entre 6 a 24 meses que ingressaram em programa de distribuição de leite fortificado \"Projeto Vivaleite\" de 2003 a 2008, em 311 municípios do Estado de São Paulo. O crescimento foi medido por meio dos valores de escore z de peso para idade (PI), calculados pelo padrão OMS/2006, obtidos, na rotina do programa, ao ingressar e a cada 4 meses durante a permanência. Os critérios de inclusão foram ter idade ao ingressar entre 6 a 24 meses, ter pelo menos duas pesagens, incluindo a obtida na entrada, e não ter relatos de problemas de saúde. As crianças foram divididas em três grupos de escore z ao entrar: sem comprometimento de peso (z> -1); risco de baixo peso (-2 &#8804;z< -1) e baixo peso (z<-2). Utilizou-se regressão linear multinível (modelo misto), permitindo a comparação, em cada idade, das médias ajustadas do escore z de ingressantes e participantes há pelo menos quatro meses, ajustadas para correlação entre medidas repetidas. Resultados: Verificou-se efeito positivo do Programa no ganho de peso das crianças, variando em função do estado nutricional ao ingressar; para as que entraram sem comprometimento de peso o ganho médio ajustado foi 0,1827 escore z, entre as que entraram com risco de baixo peso foi 0,5659 e entre as ingressantes com baixo peso foi 1,0049 escore z. Conclusões: O programa é efetivo para o crescimento infantil, medido pelo escore z PI, com efeito mais pronunciado entre as crianças que entram no programa em condições menos favoráveis de peso. / OBJECTIVE: To assess the effectiveness of a state-run food supplementation program for child growth according to childrens nutritional status at enrollment. METHODS: Cohort study including secondary data of 25,433 low-income children aged between 6 and 24 months enrolled in a fortified milk program \"Projeto Vivaleite\" in 311 cities in the State of São Paulo, Brazil, between 2003 and 2008. Children\'s growth was assessed based on weight-for-age (WA) z-scores, estimated following WHO criteria (2006). Data was routinely collected at the program enrollment and every 4 months. Inclusion criteria were: being 6 to 24 months of age at enrollment; having at least two weight measures including the first measure at enrollment; and not having any ill health conditions. At enrollment, children were categorized into three groups based on their z-scores: no compromised weight gain (z> 1); at risk of low weight (-2 &#8804;z< -1), and low weight (z< -2). Multilevel linear regression analysis (mixed model) was performed for comparison, considering age, of adjusted average z-scores between new children enrolled and those in the program for at least four months, adjusted for correlation between repeat measures. RESULTS: The program had a positive effect on children\'s weight gain. Based on their nutritional status at enrollment, adjusted average weight gain z-score was 0,1827 in children with compromised weight gain, 0,5659 in those at risk of low weight, and 1,0049 in those with low weight. CONCLUSIONS: The milk program is effective for child growth, as measured by WA z-scores. The most pronounced effect was seen among children who showed less favorable levels of weight at enrollment.