On the shrinkage of local linear curve estimators

Local linear curve estimators are typically constructed using a compactly supported kernel, which minimizes edge effects and (in the case of the Epanechnikov kernel) optimizes asymptotic performance in a mean square sense. The use of compactly supported kernels can produce numerical problems, however. A common remedy is ridging, which may be viewed as shrinkage of the local linear estimator towards the origin. In this paper we propose a general form of shrinkage, and suggest that, in practice, shrinkage be towards a proper curve estimator. For the latter we propose a local linear estimator based on an infinitely supported kernel. This approach is resistant against selection of too large a shrinkage parameter, which can impair performance when shrinkage is towards the origin. It also removes problems of numerical instability resulting from using a compactly supported kernel, and enjoys very good mean squared error properties.     

A hybrid ARIMA and support vector machines model in stock price forecasting

Traditionally, the autoregressive integrated moving average (ARIMA) model has been one of the most widely used linear models in time series forecasting. However, the ARIMA model cannot easily capture the nonlinear patterns. Support vector machines (SVMs), a novel neural network technique, have been successfully applied in solving nonlinear regression estimation problems. Therefore, this investigation proposes a hybrid methodology that exploits the unique strength of the ARIMA model and the SVMs model in forecasting stock prices problems. Real data sets of stock prices were used to examine the forecasting accuracy of the proposed model. The results of computational tests are very promising.     

Testing Inference in Inflated Beta Regressions under Model Misspecification

We consider testing inference in inflated beta regressions subject to model misspecification. In particular, quasi-z tests based on sandwich covariance matrix estimators are described and their finite sample behavior is investigated via Monte Carlo simulations. The numerical evidence shows that quasi-z testing inference can be considerably more accurate than inference made through the usual z tests, especially when there is model misspecification. Interval estimation is also considered. We also present an empirical application that uses real (not simulated) data.     

Notes on Interval Estimation of Odds Ratio in Matched Pairs under Stratified Sampling

We develop four asymptotic interval estimators and one exact interval estimator for the odds ratio (OR) under stratified random sampling with matched pairs. We apply Monte Carlo simulation to evaluate the performance of these five interval estimators. We note that the conditional score test-based interval estimator with a monotonic transformation and the interval estimator based on the Mantel–Haenszel (MH) type point estimator with the logarithmic transformation are generally preferable to the others considered here. We also note that the conditional exact confidence interval can be of use when the total number of matched pairs with discordant responses is small.     

Limit theorems for dependent Bernoulli variables with statistical inference

We consider a class of dependent Bernoulli variables where the conditional success probability is a linear combination of the last few trials and the original success probability. We obtain its limit theorems including the strong law of large numbers, weak invariance principle, and law of the iterated logarithm. We also derive some statistical inference results which make the model applicable. Simulation results are exhibited as well to show that with small sample size the convergence rate is satisfying and the proposed estimators behave well.     

Parameter Estimation of Autoregressive Models Using the Iteratively Robust Filtered Fast-τ Method

Utilizing time series modeling entails estimating the model parameters and dispersion. Classical estimators for autocorrelated observations are sensitive to presence of different types of outliers and lead to bias estimation and misinterpretation. It is important to present robust methods for parameters estimation which are not influenced by contaminations. In this article, an estimation method entitled Iteratively Robust Filtered Fast? τ(IRFFT) is proposed for general autoregressive models. In comparison to other commonly accepted methods, this method is more efficient and has lower sensitivity to contaminations due to having desirable robustness properties. This has been demonstrated by applying MSE, influence function, and breakdown point criteria.     

An Individual Mixed‐Evaluation Method for Career Intervention

Economic issues linked to career counseling are a cause for concern to policy makers in developed countries because they expect career practitioners to provide evidence of the efficiency of career counseling interventions. The aim of this study was to test an individual evaluation method mixing time series (outcomes) and life narrative (processes). The method used 5 items related to 1 client's career decision self‐efficacy and studied the evolution of those items throughout the intervention of 1 career counselor (43 days). Changepoint analysis helped in identifying the changes that have to be taken into account for time series and which are contextualized in the client's verbatim analysis. This mixed method highlighted that the career counselor's intervention increased the client's career decision self‐efficacy. Practitioners could use the methodology proposed in this article to evaluate their interventions. They could also report their practice to clients, employers, and decision makers.     

Population structure and correlation between auxiliary and target variables may affect precision of estimates in forest inventory

Simulations of forest inventory in several populations compared simple random with “quick probability proportional to size” (QPPS) sampling. The latter may be applied in the absence of a list sampling frame and/or prior measurement of the auxiliary variable. The correlation between the auxiliary and target variables required to render QPPS sampling more efficient than simple random sampling varied over the range 0.3–0.6 and was lower when sampling from populations that were skewed to the right. Two possible analytical estimators of the standard error of the estimate of the mean for QPPS sampling were found to be less reliable than bootstrapping.     

Trend smoothness achieved by penalized least squares with the smoothing parameter chosen by optimality criteria

This work presents a study about the smoothness attained by the methods more frequently used to choose the smoothing parameter in the context of splines: Cross Validation, Generalized Cross Validation, and corrected Akaike and Bayesian Information Criteria, implemented with Penalized Least Squares. It is concluded that the amount of smoothness strongly depends on the length of the series and on the type of underlying trend, while the presence of seasonality even though statistically significant is less relevant. The intrinsic variability of the series is not statistically significant and its effect is taken into account only through the smoothing parameter.     

Marginal likelihood estimation from the Metropolis output: tips and tricks for efficient implementation in generalized linear latent variable models

The marginal likelihood can be notoriously difficult to compute, and particularly so in high-dimensional problems. Chib and Jeliazkov employed the local reversibility of the Metropolis–Hastings algorithm to construct an estimator in models where full conditional densities are not available analytically. The estimator is free of distributional assumptions and is directly linked to the simulation algorithm. However, it generally requires a sequence of reduced Markov chain Monte Carlo runs which makes the method computationally demanding especially in cases when the parameter space is large. In this article, we study the implementation of this estimator on latent variable models which embed independence of the responses to the observables given the latent variables (conditional or local independence). This property is employed in the construction of a multi-block Metropolis-within-Gibbs algorithm that allows to compute the estimator in a single run, regardless of the dimensionality of the parameter space. The counterpart one-block algorithm is also considered here, by pointing out the difference between the two approaches. The paper closes with the illustration of the estimator in simulated and real-life data sets.