BACKGROUND: Parents of children with juvenile idiopathic arthritis (JIA) are often forced to make trade-offs between the effectiveness, convenience, safety, and cost of drug treatments for their child. METHODS: A discrete choice experiment was administered to parents of children with JIA to determine their preferences for drug treatments. Multinomial logit regression was used to estimate part-worth utilities and willingness-to-pay. RESULTS: Participation in daily activities was the most important attribute, followed by child reported pain. Child age, gender, years with JIA, and household income had the greatest impact on preferences. Parents’ were willing to pay $2,080 to switch from a drug representing methotrexate to etanercept (95% CI $698, $4,065). CONCLUSIONS: Parents of children with JIA have the highest maximum willingness-to-pay for drug treatments that improve daily functioning and reduce pain. Cost is a significant factor in the decisions that parents make surrounding the best treatment for a child.

Discrete choice models (DCMs) are used to forecast demand in a variety of engineering, marketing, and policy contexts, and understanding the uncertainty associated with model forecasts is crucial to inform decision-making. This thesis evaluates the suitability of DCMs for forecasting automotive demand. The entire scope of this investigation is too broad to be covered here, but I explore several elements with a focus on three themes: defining how to measure forecast accuracy, comparing model specifications and forecasting methods in terms of prediction accuracy, and comparing the implications of model specifications and forecasting methods on vehicle design. Specifically I address several questions regarding the accuracy and uncertainty of market share predictions resulting from choice of utility function and structural specification, estimation method, and data structure assumptions. I1 compare more than 9,000 models based on those used in peer-reviewed literature and academic and government studies. Firstly, I find that including more model covariates generally improves predictive accuracy, but that the form those covariates take in the utility function is less important. Secondly, better model fit correlates well with better predictive accuracy; however, the models I construct— representative of those in extant literature— exhibit substantial prediction error stemming largely from limited model fit due to unobserved attributes. Lastly, accuracy of predictions in existing markets is neither a necessary nor sufficient condition for use in design. Much of the econometrics literature on vehicle market modeling has presumed that biased coefficients make for bad models. For purely predictive purposes, the drawbacks of potentially mitigating bias using generalized method of moments estimation coupled with instrumental variables outweigh the expected benefits in the experiments conducted in this dissertation. The risk of specifying invalid instruments is high, and my results suggest that the instruments frequently used in the automotive demand literature are likely invalid. Furthermore, biased coefficients are not necessarily bad for maximizing the predictive power of the model. Bias can even aid predictions by implicitly capturing persistent unobserved effects in some circumstances. Including alternative specific constants (ASCs) in DCM utility functions improves model fit but not necessarily forecast accuracy. For frequentist estimated models all tested methods of forecasting ASCs improved share predictions of the whole midsize sedan market over excluding ASC in predictions, but only one method results in improved long term new vehicle, or entrant, forecasts. As seen in a synthetic data study, assuming an incorrect relationship between observed attributes and the ASC for forecasting risks making worse forecasts than would be made by a model that excludes ASCs entirely. Treating the ASCs as model parameters with full distributions of uncertainty via Bayesian estimation is more robust to selection of ASC forecasting method and less reliant on persistent market structures, however it comes at increased computational cost. Additionally, the best long term forecasts are made by the frequentist model that treats ASCs as calibration constants fit to the model post estimation of other parameters.

Innovation diffusion models have been studied extensively to forecast and explain the adoption process for new products or services. These models are often formulated using one of two approaches: The first, and most common is a macro-level approach that aggregates much of the market behaviour. An advantage of this method is that forecasts and other analyses may be performed with the necessity of estimating few parameters. The second is a micro-level approach that aims to utilize microeconomic information pertaining to the potential market and the innovation. The advantage of this methodology is that analyses allow for a direct understanding of how potential customers view the innovation. Nevertheless, when individuals are making adoption decisions, the reality of the situation is that the process consists of at least two stages: First, a potential adopter must become aware of the innovation; and second the aware individual must decide to adopt. Researchers, have studied multi-stage diffusion processes in the past, however a majority of these works employ a macro-level approach to model market flows. As a result, a direct understanding of how individuals value the innovation is lacking, making it impossible to utilize this information to model realistic word-of-mouth behaviour and other network dynamics. Thus, we propose a two-stage integrated model that utilizes the benefits of both the macro- and micro-level approaches. In the first stage, potential customers become aware of the innovation, which requires no decision making by the individual. As a result, we employ a macro-level diffusion process to describe the first stage. However, in the second stage potential customers decide whether to adopt the innovation or not, and we utilize a micro-level methodology to model this. We further extend the application to include forward looking behaviour, heterogeneous adopters and segmented Bayesian learning, and utilize the adopter's satisfaction levels to describe biasing and word-of-mouth behaviour. We apply the proposed model to Canadian colour-TV data, and cross-validation results suggest that the new model has excellent predictive capabilities. We also apply the two-stage model to early U.S. hybrid-electric vehicle data and results provide insightful managerial observations.

Recent interest in identifying and measuring health outcomes represents an advance in our understanding of how health care for individuals should be evaluated. However, the concept of health outcomes has mainly focussed on improvements in health status. Non-health outcomes of health care may also be important to patients. In this thesis, four tasks were undertaken with the aim of identifying non-health outcomes and establishing the extent of their relevance and importance to patients. First, the illness experience literature was reviewed to identify potential non-health outcomes. Seven categories of non-health outcomes were identified: information, being treated with dignity, being able to trust the health care provider, having distress recognised and supported, participating in decision making, legitimation and reassurance. Second, to gain an in-depth understanding of these concepts, topic-specific literature was reviewed and synthesised. Third, in order to confirm how relevant and important the concepts were to patients, a qualitative study was conducted with each of two different groups of health service users. Broadly, patients considered that all the non-health concepts were relevant, although the extent to which they were important varied. Fourth, to test the relative importance of the seven concepts, a Stated Preference Discrete Choice experiment in the context of general practice was conducted. This study showed that most people thought their GP demonstrated behaviour likely to result in the production of non-health outcomes. The results showed that although all the non-health outcomes were, to some extent, preferred by respondents, trust was most important, followed by legitimation and recognition of and support for emotional distress. Once again, these results point to the importance of context in the evaluation of health care from the patient's perspective. While still being perceived as positive aspects of health care, the provision of information and acting autonomously or participating in decisions about their health care were the non-health outcomes considered least important by patients

Discrete choice experiments (DCEs) are a stated preference valuation method. As a ubiquitous component of healthcare delivery, risk is increasingly used as an attribute in DCEs. Risk is a complex concept that is open to misinterpretation; potentially undermining the robustness of DCEs as a valuation method. This thesis employed quantitative, qualitative and eye-tracking methods to understand if and how risk communication formats affected individuals’ choices when completing a DCE and the valuations derived. This thesis used a case study focussing on the elicitation of women’s preferences for a national breast screening programme. Breast screening was chosen because of its relevance to primary care and potential contribution to the ongoing debate about the benefits and harms of mammograms. A DCE containing three attributes (probability of detecting a cancer; risk of unnecessary follow-up; and cost of screening) was designed. Women were randomised to one of two risk communication formats: i) percentages only; or ii) icon arrays and percentages (identified from a structured review of risk communication literature in health).Traditional quantitative analysis of the discrete choices made by 1,000 women recruited via an internet panel revealed the risk communication format made no difference in terms of either preferences or the consistency of choices. However, latent class analysis indicated that women’s preferences for breast screening were highly heterogeneous; with some women acquiring large non-health benefits from screening, regardless of the risks, and others expressing complete intolerance for unnecessary follow-ups, regardless of the benefits. The think-aloud method, identified as a potential method from a systematic review of qualitative research alongside DCEs, was used to reveal more about DCE respondents’ decision-making. Nineteen face-to-face cognitive interviews identified that respondents felt more engaged with the task when risk was presented with an additional icon array. Eye-tracking methods were used to understand respondents’ choice making behaviour and attention to attributes. The method was successfully used alongside a DCE and provided valid data. The results of the eye-tracking study found attributes were visually attended to by respondents most of the time. For researchers seeking to use DCEs for eliciting individuals’ preferences for benefit-risk trade-offs, respondents were more receptive to risk communicated via an icon array suggesting this format is preferable. Policy-makers should acknowledge preference heterogeneity, and its drivers, in their appraisal of the benefits of breast screening programmes. Future research is required to test alternative risk communication formats and explore the robustness of eye-tracking and qualitative research methods alongside DCEs.

Doctor of Philosophy / Department of Agricultural Economics / Jason S. Bergtold / Jessica L. Heier Stamm / Applied research in agricultural economics often involves a discrete process. Most commonly, these applications entail a conceptual framework, such as random utility, that describes a discrete-variable data-generating process. Assumptions in the conceptual framework then imply a particular empirical model. Common approaches include the binary logit and probit models and the multinomial logit when more than two outcomes are possible. Conceptual frameworks based on a discrete choice process have also been used even when the dependent variable of interest is continuous. In any case, the standard models may not be well suited to the problem at hand, as a result of either the assumptions they require or the assumptions they impose. The general theme of this dissertation is to adopt seldom-used empirical models to standard research areas in the field through applied studies. A common motivation in each paper is to lessen the exposure to specification concerns associated with more traditional models. The first paper is an attempt to provide insights into what --- if any --- weather patterns farmers respond to with respect to cropping decisions. The study region is a subset of 11 north-central Kansas counties. Empirically, this study adopts a dynamic multinomial logit with random effects approach, which may be the first use of this model with respect to farmer land-use decisions. Results suggest that field-level land-use decisions are significantly influenced by past weather, at least up to ten years. Results also suggest, however, that that short-term deviations from the longer trend can also influence land-use decisions. The second paper proposes multiple-output artificial neural networks (ANNs) as an alternative to more traditional approaches to estimating a system of acreage-share equations. To assess their viability as an alternative to traditional estimation, ANN results are compared to a linear-in-explanatory variables and parameters heteroskedastic and time-wise autoregressive seemingly unrelated regression model. Specifically, the two approaches are compared with respect to model fit and acre elasticities. Results suggest that the ANN is a viable alternative to a simple traditional model that is misspecified, as it produced plausible acre-response elasticities and outperformed the traditional model in terms of model fit. The third paper proposes ANNs as an alternative to the traditional logit model for contingent valuation analysis. With the correct network specifications, ANNs can be viewed as a traditional logistic regression where the index function has been replaced by a flexible functional form. The paper presents methods for obtaining marginal effect and willingness-to-pay (WTP) measures from ANNs, which has not been provided by the existing literature. To assess the viability of this approach, it is compared with the traditional logit and probit models as well an additional semi-nonparametric estimator with respect to model fit, marginal effects, and WTP estimates. Results suggest ANNs are viable alternative and may be preferable if misspecification of the index function is a concern.

京都大学 / 0048 / 新制・課程博士 / 博士(農学) / 甲第19031号 / 農博第2109号 / 新制||農||1031(附属図書館) / 学位論文||H27||N4913(農学部図書室) / 31982 / 京都大学大学院農学研究科生物資源経済学専攻 / (主査)教授 栗山 浩一, 教授 福井 清一, 准教授 秋津 元輝 / 学位規則第4条第1項該当 / Doctor of Agricultural Science / Kyoto University / DGAM

The primary objective of this thesis is to assess the different statistical methods used forcalibrating the Samgods model, which is a transportation planning tool employed in Sweden toforecast the demand for freight transport. By focusing on the specific context of national logisticsmodels, this research aims to enhance the accuracy and reliability of the Samgods model throughproposed improvements.In addition to evaluating the calibration techniques for the Samgods model, this thesisexplores the broader application of statistical estimation methods in national logistics models.It examines their potential benefits and limitations in order to shed light on their significance.The findings of this research highlight the crucial role of statistical estimations in improvingthe accuracy of national logistics models, thus enabling better-informed decision-making intransport planning and logistics management.By estimating the cost sensitivity parameter in the Swedish national logistics model, Samgods,this thesis contributes to a deeper understanding of the role of statistical estimations in optimizingsuch models. It underscores the importance of reliable and accurate data analysis in transportationplanning and logistics management. Ultimately, the aim is to provide valuable insights into howstatistical estimations can enhance the effectiveness of national logistics models.

<p>Blood products are used for transfusion in many routine procedures as well as emergency medical care. The balance between the supply and demand of blood products in Canada is being threatened by an increasing aging population, a growing immigrant population, and advances in medical technology which places additional strain on the blood supply. The objective of this research is to investigate the effects of demographic determinants and clinic accessibility on the frequency of blood donation in Canada excluding the province of Québec, providing a national assessment of blood donor correlates at the individual level. Exploration of these demographic factors in addition to clinic accessibility may help to explain why a limited number of repeat donors are currently contributing, with many donors giving blood only once a year. Repeat donors are vital to maintain a safe and secure blood supply, therefore it is important to retain existing donors in addition to recruiting new volunteers. In this study, individual donor and clinic information is obtained from the Canadian Blood Services 2008 national dataset, with contextual data from the 2006 Canadian Census. Discrete choice models are used to assess the effects of these variables on the frequency of blood donation across the country, highlighting the importance of clinic accessibility. The analysis is prepared for major Census Metropolitan Areas in Canada. Results may contribute to service optimization and targeted advertising, ultimately aiming to encourage the eligible population to donate.</p> / Master of Arts (MA)

This dissertation consists of three essays on forms of preference heterogeneity in discrete choice models. The first essay uses a model of heterogeneity conditional on observed individual-specific characteristics to tailor financial incentives to enhance weight loss program participation among target demographics. Financial incentives in weight loss programs have received attention mostly with respect to effectiveness rather than participation and representativeness. This essay examines the impact of financial incentives on participation with respect to populations vulnerable to obesity and understudied in the weight loss literature. We found significant heterogeneity across target sub-populations and suggest a strategy of offering multiple incentive designs to counter the dispersive effects of preference heterogeneity. The second essay investigates the ability of a novel elicitation format to reveal decision strategy heterogeneity. Attribute non-attendance, the behaviour of ignoring some attributes when performing a choice task, violates fundamental assumptions of the random utility model. However, self-reported attendance behaviour on dichotomous attendance scales has been shown to be unreliable. In this essay, we assess the ability of a polytomous attendance scale to ameliorate self-report unreliability. We find that the lowest point on the attendance scale corresponds best to non-attendance, attendance scales need be no longer than two or three points, and that the polytomous attendance scale had limited success in producing theoretically consistent results. The third essay explores available approaches to model different features of unobserved heterogeneity. Unobserved heterogeneity is popularly modelled using the mixed logit model, so called because it is a mixture of standard conditional logit models. Although the mixed logit model can, in theory, approximate any random utility model with an appropriate mixing distribution, there is little guidance on how to select such a distribution. This essay contributes to suggestions on distribution selection by describing the heterogeneity features which can be captured by established parametric mixing distributions and more recently introduced nonparametric mixing distributions, both of a discrete and continuous nature. We provide empirical illustrations of each feature in turn using simple mixing distributions which focus on the feature at hand. / Ph. D.