Dados composicionais são vetores, chamados de composições, cujos componentes são todos positivos, satisfazem a soma igual a 1 e possuem um espaço amostral próprio chamado Simplex. A restrição da soma induz a correlação entre os componentes. Isso exige que os métodos estatísticos para análise desses conjuntos de dados considerem esse fato. A teoria para dados composicionais foi desenvolvida inicialmente por Aitchison na década de 80. Desde então, várias técnicas e métodos têm sido desenvolvidos para a modelagem dos dados composicionais. Este trabalho apresenta as principais abordagens para a análise estatística de dados composicionais independentes. Sendo, regressão Dirichlet (distribuição natural aos dados composicionais) ou o uso de transformações em razões logarítmicas que saem do espaço simplex para o espaço real. Também descreve os métodos para os casos em que a suposição de independência não pode ser atendida. Por exemplo, dados composionais com dependência espacial. Para esses casos, há na literatura métodos baseados nas teorias desenvolvidas para análise geoestatística de dados univariados; ou, no uso de transformações em razões logarítmicas com a inclusão da dependência espacial. Além de revisitar os métodos já difundidos, propõe-se o uso do método de Equações de Estimação Generalizadas (EEG) como alternativa para a análise de dados composicionais independentes e com dependência espacial. A principal vantagem é que as equações de estimação necessitam apenas da especificação de funções que descrevam a média e a estrutura de covariância. Assim, não é necessário atribuir uma distribuição de probabilidade aos dados ou fazer o uso de transformações. A aplicação do método EEG para dados composicionais independentes apresentou resultados tão eficientes quanto a regressão Dirichlet ou transformação em razões logarítmicas. Para os dados composicionais com dependência espacial, o método baseado em verossimilhança foi o que apresentou valores preditos mais próximos aos valores reais. O método EEG foi mais eficaz do que a abordagem geoestatística dos componentes individuais, porém, comparado com os demais métodos, foi o que apresentou maior valor residual. / C ompositional data are vectors, called compositions, whose components are all positive, it satisfies the sum equal one and has a Simplex space. The sum constraint induces the correlation between the components and this requires that the statistical methods for the analysis of datasets consider this fact. The theory for compositional data was developed mainly by Aitchison in the 1980s, and since then, several techniques and methods have been developed for compositional data modelling. This work presents the main approaches for the statistical analysis of independent compositional data, such as Dirichlet regression (natural distribution to compositional data) or the use of transformations log-ratios that aim to leave the simplex space for to Euclidean space. Also describes the methods for cases where the assumption of independence cannot be satisfied, for example, spatial dependence compositional data. For these cases, there are in the literature methods of analysis based on the theories developed for univariate geostatistics analysis or use of logratios transformations with the inclusion of the spatial dependence generated by the distance between the points. In addition, to revisiting the already diffused methods, this work propose the use of the Generalized Estimation Equation (GEE) method as an alternative for the analysis of independent compositional data and with spatial dependence. The GEE only requires the specification of functions that describe the mean and correlation matrix (covariance structure, therefore, it is not necessary to assign a probability distribution to the data or transformations. The application of the GEE method for independent compositional data presented results as efficient as Dirichlet regression or log-ratios transformation. Compositional data with spatial dependence, log-ratios transformations presented predicted values close to the real values. GEE method was more effective than the traditional geostatistical approach, however, compared with the other methods, It was the one that presented the high residual values.

This dissertation consists of three essays studying the economics of risky health behaviors. Essay 1 estimates the effects of Graduated Driver Licensing (GDL) restrictions on weight status among adolescents aged 14 to 17 in the U.S. The findings suggest that a night curfew significantly raises adolescents’ probability of being “overweight or obese” by 1.32 percentage points, corresponding to an increase in “overweight or obesity” rate of 4.8%. A night curfew combined with a passenger restriction increases this rate by 5.8%. Overall, I estimate that nearly 16% of the rise in “overweight or obesity” rate among teenagers aged 14 to 17 in the U.S from 1999 to 2015 can be explained by the presence of the GDL restrictions. In addition, the restrictions reduce teenagers’ exercise frequency while increasing their time spent watching TV, which may help to explain the adverse effects on obesity. Essay 2 exploits the effects of the Graduated Driver Licensing (GDL) restrictions on youth smoking and drinking. It finds that being subject to minimum entry age, a learner stage, or only a night curfew has no statistically significant effect whereas, interestingly, a night curfew combined with a passenger restriction reduces youth smoking and drinking. The estimated effects become more statistically significant and larger in magnitude in the medium run, which is in line with the addictive nature of these substances. Essay 3 investigates the underlying causes of suicide. It uses data from the U.S. at the county level and the primary methodology is a two-level Bayesian hierarchical model with spatially correlated random effects. The results show that the significant effects of observable factors on suicides found by earlier research may partially stem from excluding small area effects and time trends, without controlling for which the true contribution of unobserved propensities and time trends can be hidden within observable factors. Most importantly, a lot can be learned from unobserved yet persistent propensity toward suicide captured by the spatially correlated county specific random effects. Resources should be allocated to counties with high suicide rates, but also counties with low raw suicide rates but high unobserved propensities of suicide.

We estimate the speed of income convergence for a sample of 196 European NUTS 2 regions over the period 1985-1999. So far there is no direct estimator available for dynamic panels with strong spatial dependencies. We propose a two-step procedure, which involves first spatial filtering of the variables to remove the spatial correlation, and application of standard GMM estimators for dynamic panels in a second step. Our results show that ignorance of the spatial correlation leads to potentially misleading results. Applying a system GMM estimator on the filtered variables, we obtain a speed of convergence of 6.9 per cent and a capital elasticity of 0.43. / Series: EI Working Papers / Europainstitut

Being a crucial sector in Europe’s economy, small and medium-sized enterprises (SMEs) require involvement in innovative activities to perpetuate their competitiveness. Nevertheless, European-wide funding programs that aim to foster innovation at the regional level have been criticized for not being adequately tailored to SMEs’ innovation patterns and dynamics. Against this background, this thesis sheds light on the innovation processes occurring among SMEs by explaining their innovation performance as a function of potential innovation drivers. Mindful of the relevance of geography in innovative activities, this thesis investigates whether SME innovation performance in a European region is subject to spatial dependence.   Firstly, the presence of spatial dependence is determined using Moran’s I to indicate the magnitude and significance of spatial autocorrelation in the level of SME innovation performance. In the second step, a cross-sectional spatial regression analysis examines the drivers of the innovation performance while accounting for spatial autoregressive processes. This analysis follows a bottom-up approach proposed by Elhorst (2010) to specify the suitable model for SME product and business process innovations. Additionally, the Generalized Spatial Two-stage Least Squares (GS2SLS) method accounts for heteroskedasticity of any form in the disturbances. This paper finds evidence for the presence of spatial dependence in the region’s level of SME product and business process innovation performance, implying that regions with high levels of SME innovation performance tend to be surrounded by regions with high levels and vice versa. The results from the regression analysis indicate that SMEs draw from non-R&D activities and collaboration, which offset the disadvantages these firms face. Moreover, while public and private R&D expenditures still play a role in product innovation, at least partly, the involvement in R&D activities is less important for SMEs introducing process innovation. Based on the results obtained, this paper proposes policy adaptations allowing a better environment for SMEs to participate in innovative activities.

This dissertation is comprised of three studies which examine, among other issues, racial economic disparity. The first study examines racial and gender wage gaps and considers preferences among supervisors and workers as possible sources of wage differentials. After controlling for various wage determinants, I find little statistical evidence of a racial wage gap. However, I do find evidence of a significant gender wage gap. Also, the race of an individual‟s supervisor is found to be unimportant, but workers with male supervisors are found to earn significantly higher wages than workers with female supervisors. The results reveal little evidence of employee discrimination. However, it is found that both white and male workers receive a wage premium when working for a white male supervisor. I find these results to be strongest for, and possibly driven by, small firms in the South. The second study examines the factors that Bowl Championship Series (BCS) universities use in their decision to offer athletic scholarships to high school football players. I find that a player‟s physical characteristics are important in determining the number of scholarship offers he will receive as well as his athletic performance in high school. However, a player‟s high school grade point average is not a significant determinant of the number of scholarship offers he receives. The analysis also indicates a significantly higher labor market demand for African-American high school football players, and there is also evidence of racial position segregation as well. The third study analyzes the relationship between the racial makeup of counties and economic growth and convergence in the southern U.S. The results provide strong evidence that spatial dependence is present in the data, and it is determined that the spatial lag model is appropriate for modeling the data. Significant evidence of conditional beta-convergence among the counties in the sample is found. The results also reveal that the balanced growth paths of counties are inversely related with the percentage of the county population that is African-American. That is, counties with a higher concentration of African-Americans tend to exhibit relatively slow rates of income growth.

No description available.

Doctor of Philosophy / Department of Economics / Dong Li / This dissertation is a combination of three essays on spatial econometrics and labor economics. Essays 1 and 2 developed double length regression (DLR) tests for testing functional form and spatial dependence, which includes spatial error dependence and spatial lag dependence. More specifically, these essays derive the DLR joint, DLR one-direction, and DLR conditional tests for testing functional forms and spatial dependence. The essays also provide empirical examples and Monte Carlo simulations to examine how the DLR tests perform in the empirical work and how the power of the DLR test depends on changes in functional form and spatial dependence. The results suggested that DLR tests work similarly to its Lagrangian Multiplier (LM) counterpart for testing functional form and spatial dependence in the empirical example and simulations. The DLR tests do not require the second-order derivatives of the log-likelihood function, so they provide practitioners an easy-to-use method to test for functional forms and spatial dependence. Essay 3 investigates the effects of fertility on parental labor force participation and labor supply in Vietnam. The essay uses instrumental variable (IV) probit models to estimate the effects of fertility on parental labor force participation and the IV models to estimate the effects of fertility on parental labor supply. Using the gender of the first child and the same gender of the first two children as two instrumental variables, this essay found negative effects of fertility on maternal labor force participation and labor supply. It also found positive effects of fertility on paternal labor force participation and labor supply. The results suggest that fertility had the specialization effect on parental labor force participation and labor supply in Vietnam. The homogeneity test results indicate that the magnitude of the effects of fertility on parental labor force participation and labor supply is different among parents and locations.

This paper develops an estimator for higher-order spatial autoregressive panel data error component models with spatial autoregressive disturbances, SARAR(R,S). We derive the moment conditions and optimal weighting matrix without distributional assumptions for a generalized moments (GM) estimation procedure of the spatial autoregressive parameters of the disturbance process and define a generalized two-stage least squares estimator for the regression parameters of the model. We prove consistency of the proposed estimators, derive their joint asymptotic distribution, and provide Monte Carlo evidence on their small sample performance.

This paper develops a unified framework for fixed and random effects estimation of higher-order spatial autoregressive panel data models with spatial autoregressive disturbances and heteroskedasticity of unknown form in the idiosyncratic error component. We derive the moment conditions and optimal weighting matrix without distributional assumptions for a generalized moments (GM) estimation procedure of the spatial autoregressive parameters of the disturbance process and define both a random effects and a fixed effects spatial generalized two-stage least squares estimator for the regression parameters of the model. We prove consistency of the proposed estimators and derive their joint asymptotic distribution, which is robust to heteroskedasticity of unknown form in the idiosyncratic error component. Finally, we derive a robust Hausman-test of the spatial random against the spatial fixed effects model. (authors' abstract) / Series: Department of Economics Working Paper Series

Made available in DSpace on 2017-07-10T19:25:16Z (GMT). No. of bitstreams: 1 Denise.pdf: 4576988 bytes, checksum: e7402e2569d1f12da9ffb8dcadfd665c (MD5) Previous issue date: 2012-02-16 / Soy is one of the main crops in Brazil and in the region of Cascavel / PR, where agricultural production is large, although some factors that affect productivity, monitoring and process management have been diagnosed by geostatistical models for analysis of agricultural data. Studies on the spatial variability of soil attributes associated with soybean yield, provide recommendations for doses o with varied rates, according to the maps created by spatial models. The diagnostic study on influential points is a recommended procedure for studies on spatial variability. Detecting the influential points through local influence allows measuring the changes that these points have influence on and the construction of the thematic map. This paper aims to present studies on local influence in linear spatial models considering as dependent variable soybean yield and as covariates Carbon (C), Calcium (Ca), Potassium (K), Magnesium (Mg), Manganese (Mn) and Phosphorus (P). The study on local influence is held in the response variable and the covariates using additive disturbances. The techniques of local influence diagnostics, according to the final results, were efficient in identifying outliers considered influential variables for the individual linear spatial model / A soja é uma das principais culturas agrícolas do Brasil, em particular da região de Cascavel/PR, onde a produção agrícola é grande, mas com fatores que afetam a produtividade, o monitoramento e o gerenciamento do processo, diagnosticados por modelos geoestatísticos para análise de dados agrícolas. Os estudos de variabilidade espacial dos atributos do solo, associados à produtividade da soja, possibilitam a recomendação da dosagem de insumos com taxas variadas, de acordo com os mapas construídos pelos modelos espaciais. O estudo de diagnóstico de pontos influentes é um procedimento recomendado nos estudos da variabilidade espacial. Detectar os pontos influentes, por meio da influência local, possibilita medir as alterações que esses pontos influenciam nos resultados e na construção do mapa temático. Este trabalho tem como objetivo apresentar estudos de influência local em modelos espaciais lineares, considerando como variável resposta a produtividade da soja e como covariáveis o Carbono (C), o Cálcio (Ca), o Potássio (K), o Magnésio (Mg), o Manganês (Mn) e o Fósforo (P). O estudo da influência local é realizado na variável resposta e nas covariáveis por meio de perturbações aditivas. As técnicas de influência local, de acordo com os resultados obtidos, foram eficientes na identificação de valores atípicos para as variáveis analisadas individualmente e utilizando modelo espacial linear