Universal codes from switching strategies

We discuss algorithms for combining sequential prediction strategies, a task which can be viewed as a natural generalisation of the concept of universal coding. We describe a graphical language based on Hidden Markov Models for defining prediction strategies, and we provide both existing and new models as examples. The models include efficient, parameterless models for switching between the input strategies over time, including a model for the case where switches tend to occur in clusters, and finally a new model for the scenario where the prediction strategies have a known relationship, and where jumps are typically between strongly related ones. This last model is relevant for coding time series data where parameter drift is expected. As theoretical contributions we introduce an interpolation construction that is useful in the development and analysis of new algorithms, and we establish a new sophisticated lemma for analysing the individual sequence regret of parameterised models.

Follow the leader if you can, hedge if you must

Follow-the-Leader (FTL) is an intuitive sequential prediction strategy that guarantees constant regret in the stochastic setting, but has poor performance for worst-case data. Other hedging strategies have better worst-case guarantees but may perform much worse than FTL if the data are not maximally adversarial. We introduce the FlipFlop algorithm, which is the first method that provably combines the best of both worlds. As a stepping stone for our analysis, we develop AdaHedge, which is a new way of dynamically tuning the learning rate in Hedge without using the doubling trick. AdaHedge refines a method by Cesa-Bianchi, Mansour, and Stoltz (2007), yielding improved worst-case guarantees. By interleaving AdaHedge and FTL, FlipFlop achieves regret within a constant factor of the FTL regret, without sacrificing AdaHedge’s worst-case guarantees. AdaHedge and FlipFlop do not need to know the range of the losses in advance; moreover, unlike earlier methods, both have the intuitive property that the issued weights are invariant under rescaling and translation of the losses. The losses are also allowed to be negative, in which case they may be interpreted as gains.

Paratransit demand of disabled people

This paper estimates the demand for transportation systems that are used primarily by disabled individuals. These systems are known as paratransit systems and have experienced large increases in number and average size over the past 15 years. We first use a national database and standard time series techniques to model aggregate demand. We then use a unique data set of administrative records from a paratransit system in central Virginia to estimate standard and nonstandard count models of individual demand. We conclude that most of the demand growth is from new passengers, but that predicting the growth of new passengers is very difficult. Our results also highlight the importance of incorporating autocorrelation and possible sample attrition into standard count models.

The effect of education programs on paratransit demand of people with disabilities

We describe a passenger education program to encourage responsible use of paratransit by people with disabilities. We use state-of-the-art econometric techniques to evaluate its success. We find that it has moderate effects on demand for transportation but large effects on how passengers use the transportation. In particular, passengers are more responsible about meeting the transportation at the curb rather than waiting for help inside their home. Cost-benefit analysis of the program suggests that it is a long-term worthwhile activity.

A disaggregate discrete choice model of transportation demand by elderly and disabled people in rural Virginia

This paper uses a correlated multinomial logit model and a Poisson regression model to measure the factors affecting demand for different types of transportation by elderly and disabled people in rural Virginia. The major results are: (a) A paratransit system providing door-to-door service is highly valued by transportation-handicapped people; (b) Taxis are probably a potential but inferior alternative even when subsidized; (c) Buses are a poor alternative, especially in rural areas where distances to bus stops may be long; (d) Making buses handicap-accessible would have a statistically significant but small effect on mode choice; (e) Demand is price inelastic; and (f) The total number of trips taken is insensitive to mode availability and characteristics. These results suggest that transportation-handicapped people take a limited number of trips. Those they do take are in some sense necessary (given the low elasticity with respect to mode price or availability). People will substitute away from relying upon others when appropriate transportation is available, at least to some degree. But such transportation needs to be flexible enough to meet the needs of the people involved.

Dynamic modelling of long-term care decisions

This paper describes and analyzes research on the dynamics of long-term care and the policy relevance of identifying the sources of persistence in caregiving arrangements (including the effect of dynamics on parameter estimates, implications for family welfare, parent welfare, child welfare, and cost of government programs). We discuss sources and causes of observed persistence in caregiving arrangements including inertia/state dependence (confounded by unobserved heterogeneity) and costs of changing caregivers. We comment on causes of dynamics including learning/human capital accumulation; burnout; and game-playing. We suggest how to deal with endogenous geography; dynamics in discrete and continuous choices; and equilibrium issues (multiple equilibria, dynamic equilibria). We also present an overview of commonly used longitudinal data sets and evaluate their relative advantages/disadvantages. We also discuss other data issues related to noisy measures of wealth and family structure. Finally, we suggest some methods to handle econometric problems such as endogeneous geography. © 2014 Springer Science+Business Media New York.

Two dynamic discrete choice estimation problems and simulation method solutions

This paper considers two problems that frequently arise in dynamic discrete choice problems but have not received much attention with regard to simulation methods. The first problem is how to simulate unbiased simulators of probabilities conditional on past history. The second is simulating a discrete transition probability model when the underlying dependent variable is really continuous. Both methods work well relative to reasonable alternatives in the application discussed. However, in both cases, for this application, simpler methods also provide reasonably good results.

The educational pipeline for health care professionals : understanding the source of racial differences

The underrepresentation of blacks in the healthcare professions may have direct implications for the health outcomes of minority patients, underscoring the importance of understanding movement through the educational pipeline into professional healthcare careers by race. We jointly model individuals' postsecondary decisions including enrollment, college type, degree completion, and choosing a healthcare occupation requiring an advanced degree. We estimate the parameters of the model with maximum likelihood using data from the NLS-72. Our results emphasize the importance of pre-collegiate factors and of jointly examining the full chain of educational decisions in understanding the sources of racial disparities in professional healthcare occupations.

Shared caregiving responsibilities of adult siblings with elderly parents

This paper uses a nonstructural, ordered discrete choice model to measure the effects of various parent and child characteristics upon the independent caregiving decisions of the adult children of elderly parents sampled in the 1982 and 1984 National Long Term Care Survey (NLTCS). While significant effects are noted, emphasis is placed on test statistics constructed to measure the independence of caregiving decisions. The test statistic results are conclusive: The caregiving decisions of adult children are dependent across time and family members. Structural models taking dependencies among family members into account note effects similar to those in the nonstructural model.

Simulation-based estimation

In this paper, I present a number of leading examples in the empirical literature that use simulation-based estimation methods. For each example, I describe the model, why simulation is needed, and how to simulate the relevant object. There is a section on simulation methods and another on simulations-based estimation methods. The paper concludes by considering the significance of each of the examples discussed a commenting on potential future areas of interest.

Strategic play among family members when making long-term care decisions

This paper describes a strategic model of bargaining within a family to determine how to care for an elderly parent. We estimate the parameters of the model using data from the National Long-term Care Survey. We find that the parameter estimates generally make sense and that the model is consistent with the data. The results have strong implications for using less structural empirical models for policy analysis.

Semiparametric estimates of the supply and demand effects of disability on labor force participation

This paper modifies and uses the semiparametric methods of Ichimura and Lee (1991) on standard cross-section data to decompose the effect of disability on labor force participation into a demand and a supply effect. It shows that straightforward use of Ichimura and Lee leads to meaningless results while imposing monotonicity on the unknown function leads to substantial results. The paper finds that supply effects dominate the demand effects of disability.

Semiparametric estimation of count regression models

This paper develops a semiparametric estimation approach for mixed count regression models based on series expansion for the unknown density of the unobserved heterogeneity. We use the generalized Laguerre series expansion around a gamma baseline density to model unobserved heterogeneity in a Poisson mixture model. We establish the consistency of the estimator and present a computational strategy to implement the proposed estimation techniques in the standard count model as well as in truncated, censored, and zero-inflated count regression models. Monte Carlo evidence shows that the finite sample behavior of the estimator is quite good. The paper applies the method to a model of individual shopping behavior. © 1999 Elsevier Science S.A. All rights reserved.

Feasibility of using technology to disseminate evidence to rural nurses and improve patient outcomes

Background: Rural African American women receive less frequent mammography screening and die of breast cancer at a higher rate than is seen in the general population. To overcome this disparity, it is necessary to assist rural providers in their efforts to influence women to obtain screening. Method: This study examined the feasibility of using distance education to disseminate knowledge about timely and appropriate mammography screening to rural nurses, using patient outcome data to evaluate the effectiveness of this intervention. Results: Overall, there was a decline in referrals and mammography screening, but the intervention group centers showed a smaller decline after the educational intervention than did the control group. Conclusion: The findings show the effect of dissemination of information and the feasibility of using patient outcome data for educational evaluation. Neighboring academic health centers and nursing schools should include in their mission the provision of educational programs for relatively isolated rural nurses.