Adaptive model selection using empirical complexities

Given $n$ independent replicates of a jointly distributed pair $(X,Y)\in {\cal R}^d \times {\cal R}$, we wish to select from a fixed sequence of model classes ${\cal F}_1, {\cal F}_2, \ldots$ a deterministic prediction rule $f: {\cal R}^d \to {\cal R}$ whose risk is small. We investigate the possibility of empirically assessingthe {\em complexity} of each model class, that is, the actual difficulty of the estimation problem within each class. The estimated complexities are in turn used to define an adaptive model selection procedure, which is based on complexity penalized empirical risk.The available data are divided into two parts. The first is used to form an empirical cover of each model class, and the second is used to select a candidate rule from each cover based on empirical risk. The covering radii are determined empirically to optimize a tight upper bound on the estimation error. An estimate is chosen from the list of candidates in order to minimize the sum of class complexity and empirical risk. A distinguishing feature of the approach is that the complexity of each model class is assessed empirically, based on the size of its empirical cover.Finite sample performance bounds are established for the estimates, and these bounds are applied to several non-parametric estimation problems. The estimates are shown to achieve a favorable tradeoff between approximation and estimation error, and to perform as well as if the distribution-dependent complexities of the model classes were known beforehand. In addition, it is shown that the estimate can be consistent,and even possess near optimal rates of convergence, when each model class has an infinite VC or pseudo dimension.For regression estimation with squared loss we modify our estimate to achieve a faster rate of convergence.

Tobin meets Oates: Solidarity and the optimal fiscal federal structure

We explore the implications for the optimal degree of fiscal decentralization when people spreferences for goods and services, which classic treatments of fiscal federalism (Oates, 1972)place in the purview of local governments, exhibit specific egalitarianism (Tobin, 1970), orsolidarity. We find that a system in which the central government provides a common minimumlevel of the publicly provided good, and local governments are allowed to use their ownresources to provide an even higher local level, performs better from an efficiency perspectiverelative to all other systems analyzed for a relevant range of preferences over solidarity.

Resting metabolic rate in six- to ten-year-old obese and nonobese children.

The resting metabolic rate (RMR) and body composition of 130 obese and nonobese prepubertal children, aged 6 to 10 years, were assessed by indirect calorimetry and skin-fold thickness, respectively. The mean (+/- SD) RMR was 4619 +/- 449 kJ.day-1 (164 +/- 31 kJ.kg body weight-1 x day-1) in the 62 boys and 4449 +/- 520 kJ.day-1 (147 +/- 32 kJ.kg body weight-1 x day-1) in the 68 girls. Fat-free mass was the best single predictor of RMR (R2 = 0.64; p < 0.001). Step-down multiple regression analysis, with independent variables such as age, gender, weight, and height, allowed several RMR predictive equations to be developed. An equation for boys is as follows: RMR (kJ.day-1) = 1287 + 28.6 x Weight(kg) + 23.6 x Height(cm) - 69.1 x Age(yr) (R2 = 0.58; p < 0.001). An equation for girls is as follows: RMR (kJ.day-1 = 1552 + 35.8 x Weight (kg) + 15.6 x Height (cm) - 36.3 x Age (yr) (R2 = 0.69; p < 0.001). Comparison between the measured RMR and that predicted by currently used formulas showed that most of these equations tended to overestimate the RMR of both genders, especially in overweight children.

The effect of treatment with calcitonin on vertebral fracture rate in osteoporosis.

The efficacy of treatments for osteoporosis does not become evident when evaluated by fracture incidence (FI). Vertebral FI decreased in all controlled studies on calcitonin, but not significantly. Small sample sizes and short periods of treatment may have masked a possible therapeutic benefit, but longer, controlled studies with sodium fluoride or etidronate in larger groups of patients also failed to show a decrease in FI. The present analysis of nine published, therapeutic studies which indicate the FI per year and the initial prevalence of vertebral fractures, examines the question of whether the initial prevalence of fractures has an effect on the subsequent incidence of new fractures and whether the therapeutic effects have to be evaluated as a function of the initial prevalence of fractures. Bearing in mind the differences in roentgenological evaluation and in the size and quality of the various studies, the analysis revealed (1) that in the control groups there was a higher FI in patients with more than three vertebral fractures at baseline (estimated odds ratio (OR) = 49, p = 0.011); (2) that a similar trend, although not statistically significant, was observed in treated patients; (3) that the groups of control patients treated for more than 1 year showed in general an increase in FI beyond the first year and that the reverse was true in treated patients. In conclusion, failure to allow for the initial prevalence of vertebral fractures at the individual level in therapeutic trials of calcitonin to treat osteoporosis and prevent new fractures might have contributed to the absence of a demonstrable benefit of the treatment in those studies.

Constant interest rate projections without the curse of indeterminacy

Constant interest rate (CIR) projections are often criticized on the grounds that they are inconsistent with the existence of a unique equilibrium in a variety of forward-looking models. This note shows howto construct CIR projections that are not subject to that criticism, using a standard New Keynesian model as a reference framework.

Optimal taxation without state-contingent debt

To recover a version of Barro's (1979) `random walk'tax smoothing outcome, we modify Lucas and Stokey's (1983) economyto permit only risk--free debt. This imparts near unit root like behaviorto government debt, independently of the government expenditureprocess, a realistic outcome in the spirit of Barro's. We showhow the risk--free--debt--only economy confronts the Ramsey plannerwith additional constraints on equilibrium allocations thattake the form of a sequence of measurability conditions.We solve the Ramsey problem by formulating it in terms of a Lagrangian,and applying a Parameterized Expectations Algorithm tothe associated first--order conditions. The first--order conditions andnumerical impulse response functions partially affirmBarro's random walk outcome. Though the behaviors oftax rates, government surpluses, and government debts differ, allocationsare very close for computed Ramsey policies across incomplete and completemarkets economies.

Income maintenance programs and multidimensional screening

This paper examines properties of optimal poverty assistance programs under different informational environments using an income maintenanceframework. To that end, we make both the income generating ability andthe disutility of labor of individuals unobservable, and compare theresulting benefit schedules with those of programs found in the UnitedStates since Welfare Reform (1996). We find that optimal programs closelyresemble a Negative Income Tax with a Benefit Reduction rate that dependson the distribution of population characteristics. A policy of workfare(unpaid public sector work) is inefficient when disutility of labor isunobservable, but minimum work requirements (for paid work) may be usedin that same environment. The distortions to work incentives and thepresence of minimum work requirements depend on the observability andrelative importance of the population's characteristics.

The economics of organized crime and optimal law enforcement

This paper extends the optimal law enforcement literature to organized crime.We model the criminal organization as a vertical structure where the principal extracts some rents from the agents through extortion. Depending on the principal's information set, threats may or may not be credible. As long as threats are credible, the principal is able to fully extract rents.In that case, the results obtained by applying standard theory of optimal law enforcement are robust: we argue for a tougher policy. However, when threats are not credible, the principal is not able to fully extract rents and there is violence. Moreover, we show that it is not necessarily true that a tougher law enforcement policy should be chosen when in presence of organized crime.

Rate of development of forensically-important Dipterain southern Brazil

Dipteran larvae were collected from rabbit (Oryctolagus cunniculus L.) carcasses during the four seasons in 2005 in the southernmost state of Rio Grande do Sul, Brazil. The larvae were fed ground beef at ambient temperatures following collection from carcasses. The development of each species under these conditions was estimated. The most abundant species in the carcasses were Lucilia eximia (Wiedemann) and Chrysomya albiceps (Wiedemann) (Calliphoridae), and they were found in all seasons. The data were fitted to a linear model that describes the relationship between temperature and linear developmental rating. These two species are primary forensic indicators in southern Brazil. Other species such as Hemilucilia semidiaphana (Rondani) (Calliphoridae), Synthesiomyia nudiseta (Wulp), Muscina stabulans (Fallen) (Muscidae), and Fannia pusio (Wiedemann) (Fanniidae) were forensically less important because they only occurred in high frequency in certain seasons and during the first days of carcass decomposition.

Optimal law enforcement and criminal organization

In this paper, we take an organizational view of organized crime. In particular, we study the organizational consequences of product illegality attending at the following characteristics: (i) contracts are not enforceable in court, (ii) all participants are subject to the risk of being punished, (iii) employees present a major threat to the entrepreneur having the most detailed knowledge concerning participation, (iv) separation between ownership and management is difficult because record-keeping and auditing augments criminal evidence.

Organizational capital and employment fluctuations

In this paper I present a model in which production requires two types of labor inputs: regular productive tasks and organizational capital, which is accumulated by workers performing organizational tasks. By allocating more workers from organizational to productive tasks, firms can temporarily increase production without hiring. The availability of this intensive margin of labor adjustment, in combination with adjustment costs along the extensive margin (search frictions, firing costs, training costs), makes it optimal to delay employment adjustments. Simulations indicate that this mechanism is quantitatively important even if only a small fraction of workers perform organizational tasks, and explains why the hiring rate is persistent and why employment is slow to recover after the end of a recession.

The new Keynesian approach to monetary policy analysis: Lessons and new directions

I discuss several lessons regarding the design and conduct of monetary policy that have emerged out of the New Keynesian research program. Those lessons include the bene.ts of price stability, the gains from commitment about future policies, the importance of nat-ural variables as benchmarks for policy, and the bene.ts of a credible anti-inflationary stance. I also point to one challenge facing NK modelling efforts: the need to come up with relevant sources of policy tradeoffs. A potentially useful approach to meeting that challenge, based on the introduction of real imperfections, is presented.

The rate-limiting step for glucose transport into the hypothalamus is across the blood-hypothalamus interface.

Specialized glucosensing neurons are present in the hypothalamus, some of which neighbor the median eminence, where the blood-brain barrier has been reported leaky. A leaky blood-brain barrier implies high tissue glucose levels and obviates a role for endothelial glucose transporters in the control of hypothalamic glucose concentration, important in understanding the mechanisms of glucose sensing We therefore addressed the question of blood-brain barrier integrity at the hypothalamus for glucose transport by examining the brain tissue-to-plasma glucose ratio in the hypothalamus relative to other brain regions. We also examined glycogenolysis in hypothalamus because its occurrence is unlikely in the potential absence of a hypothalamus-blood interface. Across all regions the concentration of glucose was comparable at a given plasma glucose concentration and was a near linear function of plasma glucose. At steady-state, hypothalamic glucose concentration was similar to the extracellular hypothalamic glucose concentration reported by others. Hypothalamic glycogen fell at a rate of approximately 1.5 micromol/g/h and remained present in substantial amounts. We conclude for the hypothalamus, a putative primary site of brain glucose sensing that: the rate-limiting step for glucose transport into brain cells is at the blood-hypothalamus interface, and that glycogenolysis is consistent with a substantial blood -to- intracellular glucose concentration gradient.

Conglomeration with bankruptcy costs: Separate or joint financing?

The paper analyzes the determinants of the optimal scope of incorporation in the presenceof bankruptcy costs. Bankruptcy costs alone generate a non-trivial tradeoff between thebenefit of coinsurance and the cost of risk contamination associated to joint financing corporate projects through debt. This tradeoff is characterized for projects with binary returns,depending on the distributional characteristics of returns (mean, variability, skewness, heterogeneity, correlation, and number of projects), the bankruptcy recovery rate, and the taxrate advantage of debt relative to equity. Our testable predictions are broadly consistentwith existing empirical evidence on conglomerate mergers, spin-offs, project finance, andsecuritization.

Physiological state and rate protein synthesis of phytoplankton off the coast of Peru as measured by sulfur-35 incorporation

Analiza el estado de la fisiología del fitoplancton de las aguas costeras cercanas a Perú