No description available.

<p>  This dissertation has two essays that are focused on understanding the effects of technology adoption decisions made by smallholder farmers who have incomplete information. The first essay employed a clustered randomized control trial (RCT) with factorial design in upper Eastern Kenya to estimate the impact of three different interventions at improving credence attributes of smallholder farmers’ maize. This essay also utilized a Becker DeGroot Marschak auction method to determine if farmers were willing to adopt a credence technology, and if yes, if their willingness to pay varied based on having previous experience with this agricultural technology. The second essay used the 2018/19 Ethiopia Socio-economic Survey to analyze the impacts of three sources of measurement error caused by farmers’ misperceptions on maize yields. Moreover, this essay explored how farmers’ incomplete information about adoption of one agricultural input led to misallocation of other complementary inputs. </p>

Regulatory Agencies are responsible for monitoring the performance of particular measurement communities. In order to achieve their objectives, they sponsor Intercomparison exercises between the members of these communities. The Intercomparison Exercise Program for Organic Contaminants in the Marine Environment is an ongoing NIST/NOAA program. It was started in 1986 and there have been 19 studies to date. Using this data as a motivation we review the theory and practices applied to its analysis. It is a common practice to apply some kind of filter to the comparison study data. These filters go from outliers detection and exclusion to exclusion of the entire data from a participant when its measurements are very “different". When the measurements are not so “different" the usual assumption is that the laboratories are unbiased then the simple mean, the weighted mean or the one way random effects model are applied to obtain estimates of the true value. Instead we explore methods to analyze these data under weaker assumptions and apply them to some of the available data. More specifically we explore estimation of models assessing the laboratories performance and way to use those fitted models in estimating a consensus value for new study material. This is done in various ways starting with models that allow a separate bias for each lab with each compound at each point in time and then considering generalizations of that. This is done first by exploiting models where, for a particular compound, the bias may be shared over labs or over time and then by modeling systematic biases (which depend on the concentration) by combining data from different labs. As seen in the analyses, the latter models may be more realistic. Due to uncertainty in the certified reference material analyzing systematic biases leads to a measurement error in linear regression problem. This work has two differences from the standard work in this area. First, it allows heterogeneity in the material being delivered to the lab, whether it be control or study material. Secondly, we make use of Fieller's method for estimation which has not been used in the context before, although others have suggested it. One challenge in using Fieller's method is that explicit expressions for the variance and covariance of the sample variance and covariance of independent but non-identically distributed random variables are needed. These are developed. Simulations are used to compare the performance of moment/Wald, Fieller and bootstrap methods for getting confidence intervals for the slope in the measurement model. These suggest that the Fieller's method performs better than the bootstrap technique. We also explore four estimators for the variance of the error in the equation in this context and determine that the estimator based on the modified squared residuals outperforms the others. Homogeneity is a desirable property in control and study samples. Special experiments with nested designs must be conducted for homogeneity analysis and assessment purposes. However, simulation shows that heterogeneity has low impact on the performance of the studied estimators. This work shows that a biased but consistent estimator for the heterogeneity variance can be obtained from the current experimental design.

In ecological population management, years of animal counts are fit to nonlinear, dynamic models (e.g. the Ricker model) because the values of the parameters are of interest. The yearly counts are subject to measurement error, which inevitably leads to biased estimates and adversely affects inference if ignored. In the literature, often convenient distribution assumptions are imposed, readily available estimated measurement error variances are not utilized, or the measurement error is ignored entirely. In this thesis, ways to estimate the parameters of the Ricker model and perform inference while accounting for measurement error are investigated where distribution assumptions are minimized and estimated measurement error variances are utilized. To these ends, SIMEX and modified estimating equations (MEE) rather than likelihood methods are investigated for data on the abundance and log-abundance scales, and how inference is done via the parametric bootstrap and estimated standard errors from the modified estimating equations is shown. Subsequently, simulation studies are performed on the log-abundance scale under varying parameter values to learn how levels of measurement error variances (ranging from the realistically low value of 0.0025 to unrealistically high value of 0.025 ) affects the estimators and inference when measurement error is ignored, and how the methods perform accounting for it. It was found that the bias induced by measurement error depends on the true value of the parameter. Furthermore, the performances of SIMEX and MEE are associated with the true value of a and the level of measurement error variance. In particular, both methods perform best for a > 1 and low to moderate levels of measurement error variance, with the MEE estimators having high standard error and often poorer performance than those from SIMEX. It was also found that the MEE estimators contain singularities which attribute to its low precision and erratic behavior. These methods were then applied to actual moose count data with sample size more than double that of the simulations. It was found that both the SIMEX and MEE estimators performed well suggesting that sample size contributes to previous poor behavior.

In this dissertation, I study econometric issues in network and health economics. Measurement error is a ubiquitous problem in the peer effects literature that is not well understood. In Chapter 1, ``Measurement error in peer effects,'' I develop a constructive approach to empirically assess the bias caused by links missing at random. I apply my method to study the bias in peer effect estimates of recreational and physical activities among adolescents in the United States. I find that the magnitude and direction of the bias depends on the estimator. Estimators that measure the aggregate effect of peers' outcomes are more robust to measurement error and can be unbiased even when fifty percent of peer interactions are unobserved. Estimators that measure the average effect of peers'’ outcomes are more susceptible to measurement error and suffer from a persistent downward bias. My findings illustrate the importance of understanding measurement error's impact, when measurement error will likely bias results and when it can be safely ignored. In Chapter 2, ``Non-random errors in peer effects,'' I study the effects of measurement error on a generalized peer effect model that nests two of the most commonly used estimators. Measurement error in the specification of peer groups leads to biased estimates. I adapt Monte Carlo methods developed for studying measurement error when peers' interactions are missing at random to understand the effects of top-coding, non-random errors and spurious peer interactions. I find that non-random errors pose the greatest threat, often leading to overestimation and persistent biases. Top-coding can also severely bias estimates when the constraint impacts a majority of individuals but otherwise has a mild effect. While spurious links in limited quantities can often be ignored. Chapter 3, ``Nurse outreach and frequent emergency department users: A synthetic control analysis,'' studies the effects of an intervention to promote primary care engagement among frequent emergency department users. Emergency departments are one of the costliest places to receive care and are routinely overcrowded. Various policy initiatives have yielded mixed findings. I use synthetic control methods to analyze the effects of a nurse outreach program for frequent emergency department users implemented by a major U.S. insurer. The program seeks to reduce emergency department utilization by promoting primary care engagement. I leverage a unique commercial claims data set to measure the effects of the program on primary care and emergency department utilization. My findings suggest that six months after treatment nurse outreach increased primary care utilization by 15 percent; however, I find no clear effect on emergency department utilization. My findings indicate that increasing primary care engagement may not be sufficient to prevent emergency department over utilization. / Economics

No description available.

No description available.

This dissertation consists of two chapters. The first chapter shows that the measurement errors in betas for stocks induce corresponding measurement errors in alphas and a spurious negative covariance between the estimated betas and alphas across stocks. This negative covariance between the estimated betas and alphas results in a violation of the independence assumption between the independent variable (betas) and error terms in the Fama-MacBeth regressions of tests of the CAPM, thereby creating a downward bias in the estimated market risk premiums. The procedure of using portfolio returns and betas does not necessarily eliminate this bias. Depending upon the grouping variable used to form portfolios, the negative covariance between estimated betas and alphas can be increased, decreased, and can even be made positive. This paper proposes two methods for correcting the downward bias in the estimated market risk premium. The estimated market risk premiums are consistent with the CAPM after the proposed corrections. The second chapter provides evidence that when the ex-post market risk premium is positive (up markets), the relation between returns and betas is positive, significant, and consistent with the CAPM. However, when the ex-post market risk premium is negative (down markets), the negative relation between betas and returns is significant, but stronger than what is implied by the CAPM. This strong negative relation offsets the positive relation, resulting in an insignificant relation between returns and betas for the overall period. The negative relation between size and returns, after controlling for beta differences, is present only when the ex-post market risk premium is negative, and is responsible for the negative relation for the overall period. This paper decomposes the negative relation between size and returns after controlling for beta differences into the intercept size effect (relation between alphas of stocks and their size) and the residual size effect (relation between residuals of stocks and their size). The asymmetrical size effect between up and down market is being driven by the residual size effect. Long term mean reversion in returns explains, in part, the negative relation between size and returns during down markets. / Ph. D.

Em modelos de regressão em que uma covariável é medida com erro, é comum o uso de estruturas que relacionam a covariável observada com a verdadeira covariável não observada. Essas estruturas são usualmente aditivas ou multiplicativas. Na literatura existem diversos trabalhos interessantes que tratam de modelos de regressão com erro de medida aditivo, muitos dos quais são modelos lineares com covariáveis e erro de medida normalmente distribuídos. Para modelos em que o erro de medida é multipicativo, não se encontra na literatura o mesmo desenvolvimento teórico encontrado para modelos em que o erro de medida é aditivo. O mesmo vale para situações em que as suposições de normalidade para as covariáveis e erro de medida não se aplicam. Este trabalho propõe a construção, definição, métodos de estimação e análise de diagnóstico para modelos de regressão com erro de medida multiplicativo em uma das covariáveis. Para esses modelos, consideramos que a variável resposta possa pertencer ou à classe de modelos de regressão série de potências modificadas ou à família exponencial. O rol de distribuições pertencentes à família série de potências modificada é bem abrangente, portanto, neste trabalho, desenvolvemos a teoria de estimação e validação do modelo primeiramente de forma geral e, para exemplificar, apresentamos o modelo de regressão binomial negativa com erro de medida. para o caso em que a variável resposta pertença à família exponencial. apresentamos o modelo de regressão beta com erro de medida multiplicativo. Todos os modelos propostos foram analisados através de estados de simulação e aplicados a conjuntos de dados reais. / In regression models in which a covariate is measured with erros, it is common to use structures that correlate the observed covariate with the true non-observed covariate. Such structures are usually additive or multiplicative. In the literatue there are several interesting works that deal with regression models having an additive measuremsnt error, many of which are linear models with covariate and measurement error normally distributed. For models having a multiplicative measurement error, one does not find in the literature the same theoretical amount of works as one finds for models in which the measurement error is additive. The same happens in situations where the supositions of normality for the covariates and the measurement errors do not apply. The presente work proposes the construction,definition, estimation methods, and diagnostic analysis for the regression models with a multiplicative measurement error in one of the covariates. For these models it is considered that the response variable may belong either to the class of modified power series regression models or to the exponential family. The list of distributions belonging to the family modified power series is rather comprehensive; for this reason this work develops, firstly and in a general way, the models estimation and validation theory, and, as an example, presents the model of negative binomial regression with measurement error. In the case where response variable belongs to the exponential family, the model of beta regression with multiplicative measurement error is presented. All proposed models were analysed through simulationb studies and applied to real data sets.

Dans les études épidémiologiques, les erreurs de mesure de l’exposition étudiée peuvent biaiser l’estimation des risques liés à cette exposition. Un grand nombre de méthodes de correction de l’effet de ces erreurs a été développé mais en pratique elles ont été rarement appliquées, probablement à cause du fait que leur capacité de correction et leur mise en œuvre sont peu maîtrisées. Une autre raison non moins importante est que, en l’absence de données répétées ou de données de validation, ces méthodes de correction exigent la connaissance détaillée des caractéristiques (taille, nature, structure et distribution) des erreurs de mesure. L’objectif principal de cette thèse est d’étudier l’impact de la prise en compte des erreurs de mesure dans les analyses du risque de décès par cancer du poumon associé à l’exposition au radon à partir de la cohorte française des mineurs d’uranium (qui ne dispose ni de données répétées, ni de données de validation). Les objectifs spécifiques étaient (1) de caractériser les erreurs de mesure associées aux expositions radiologiques (radon et ses descendants, poussières d’uranium et rayonnements gamma), (2) d’étudier l’impact des erreurs de mesure de l’exposition au radon et à ses descendants sur l’estimation de l’excès de risque relatif (ERR) de décès par cancer du poumon et (3) d’étudier et comparer la performance des méthodes de correction de l’effet de ces erreurs. La cohorte française des mineurs d’uranium comprend plus de 5000 individus exposés de manière chronique au radon et à ses descendants qui ont été suivis en moyenne pendant 30 ans. Les erreurs de mesure ont été caractérisées en prenant en compte l’évolution des méthodes d’extraction et de la surveillance radiologique des mineurs au fil du temps. Une étude de simulation basée sur la cohorte française des mineurs d’uranium a été mise en place pour étudier l’impact de ces erreurs sur l’ERR ainsi que pour comparer la performance des méthodes de correction. Les résultats montrent que les erreurs de mesure de l’exposition au radon et à ses descendants ont diminué au fil des années. Pour les premières années, avant 1970, elles dépassaient 45 % et après 1980 elles étaient de l’ordre de 10 %. La nature de ces erreurs a aussi changé au cours du temps ; les erreurs essentiellement de nature Berkson ont fait place à des erreurs de nature classique après la mise en place des dosimètres individuels à partir de 1983. Les résultats de l’étude de simulation ont montré que les erreurs de mesure conduisent à une atténuation de l’ERR vers la valeur nulle, avec un biais important de l’ordre de 60 %. Les trois méthodes de correction d’erreurs considérées ont permis une réduction notable mais partielle du biais d’atténuation. Un avantage semble exister pour la méthode de simulation extrapolation (SIMEX) dans notre contexte, cependant, les performances des trois méthodes de correction sont fortement tributaires de la détermination précise des caractéristiques des erreurs de mesure.Ce travail illustre l’importance de l’effet des erreurs de mesure sur les estimations de la relation entre l’exposition au radon et le risque de décès par cancer du poumon. L’obtention d’estimation de risque pour laquelle l’effet des erreurs de mesure est corrigé devrait s’avérer d’un intérêt majeur en support des politiques de protection contre le radon en radioprotection et en santé publique. / In epidemiological studies, measurement errors in exposure can substantially bias the estimation of the risk associated to exposure. A broad variety of methods for measurement error correction has been developed, but they have been rarely applied in practice, probably because their ability to correct measurement error effects and their implementation are poorly understood. Another important reason is that many of the proposed correction methods require to know measurement errors characteristics (size, nature, structure and distribution).The aim of this thesis is to take into account measurement error in the analysis of risk of lung cancer death associated to radon exposure based on the French cohort of uranium miners. The mains stages were (1) to assess the characteristics (size, nature, structure and distribution) of measurement error in the French uranium miners cohort, (2) to investigate the impact of measurement error in radon exposure on the estimated excess relative risk (ERR) of lung cancer death associated to radon exposure, and (3) to compare the performance of methods for correction of these measurement error effects.The French cohort of uranium miners includes more than 5000 miners chronically exposed to radon with a follow-up duration of 30 years. Measurement errors have been characterized taking into account the evolution of uranium extraction methods and of radiation protection measures over time. A simulation study based on the French cohort of uranium miners has been carried out to investigate the effects of these measurement errors on the estimated ERR and to assess the performance of different methods for correcting these effects.Measurement error associated to radon exposure decreased over time, from more than 45% in the early 70’s to about 10% in the late 80’s. Its nature also changed over time from mostly Berkson to classical type from 1983. Simulation results showed that measurement error leads to an attenuation of the ERR towards the null, with substantial bias on ERR estimates in the order of 60%. All three error-correction methods allowed a noticeable but partial reduction of the attenuation bias. An advantage was observed for the simulation-extrapolation method (SIMEX) in our context, but the performance of the three correction methods highly depended on the accurate determination of the characteristics of measurement error.This work illustrates the importance of measurement error correction in order to obtain reliable estimates of the exposure-risk relationship between radon and lung cancer. Corrected risk estimates should prove of great interest in the elaboration of protection policies against radon in radioprotection and in public health.