Flexible estimation of serial correlation in nonlinear mixed models

In the conventional linear mixed-effects model, four structures can be distinguished: fixed effects, random effects, measurement error and serial correlation. The latter captures the phenomenon that the correlation structure within a subject depends on the time lag between two measurements. While the general linear mixed model is rather flexible, the need has arisen to further increase flexibility. In addition to work done in the area, we propose the use of spline-based modeling of the serial correlation function, so as to allow for additional flexibility. This approach is applied to data from a pre-clinical experiment in dementia which studied the eating and drinking behavior in mice.     

Comment on “modeling asset returns with alternative stable distributions”

This paper considers a modified CUSUM test, suggested by Dufour (1982) for parameter instability and structural change with an unknown change point in a linear model with serially correlated disturbances, in which a preliminary estimate of the autoregressive coefficient for the error process is obtained, and used to transform the data. Then the standard CUSUM statistic is calculated on the transformed data. This paper derives the asymptotic distribution of the modified CUSUM test. We show that the modified CUSUM test retains its asymptotic significance level, i.e., the modified CUSUM test has the same asymptotic distribution as the CUSUM test with serially uncorrelated errors.     

Lagrance-multiplier tersts for weak exogeneity: a synthesis

This paper unifies two seemingly separate approaches to test weak exogeneity in dynamic regression models with Lagrange-mulptiplier statistics. The first class of tests focuses on the orthogonality between innovations and conditioning variables, and thus is related to the Durbin-Wu-Hausman specification test. The second approach has been developed more recently in the context of context of cointegration and error correction models, ad concentrates on the question whether the conditioning variables display error correction behaviour. It is shown that the vital difference between the two approaches stems from the choice of the parmeters of interest. A new test is derived, which encompasses both its predecessors. The test is applied to an error correction model of the demand for money in Switzerland.     

Nesting Monte Carlo EM for high-dimensional item factor analysis

The item factor analysis model for investigating multidimensional latent spaces has proved to be useful. Parameter estimation in this model requires computationally demanding high-dimensional integrations. While several approaches to approximate such integrations have been proposed, they suffer various computational difficulties. This paper proposes a Nesting Monte Carlo Expectation-Maximization (MCEM) algorithm for item factor analysis with binary data. Simulation studies and a real data example suggest that the Nesting MCEM approach can significantly improve computational efficiency while also enjoying the good properties of stable convergence and easy implementation.     

Breaking alias chains in fractional factorials

Some experimenters carry out their investigation in stages. They begin with an initial 2^n-p fraction of resolution IV, in which the main effects are clean and the interactions are aliased in chains, Then, having analyzed the initial experiment, they plan further runs to isolate certain interactions by breaking the chains. In this paper a method called semifolding, for choosing the points in the second experiment, is presented.     

Uniformly best estimation in linear regression when prior information is fuzzy

Modeling prior information as a fuzzy set and using Zadeh’s extension principle, a general approach is presented how to rate linear affine estimators in linear regression. This general approach is applied to fuzzy prior information sets given by ellipsoidal α-cuts. Here, in an important and meaningful subclass, a uniformly best linear affine estimator can be determined explicitly. Surprisingly, such a uniformly best linear affine estimator is optimal with respect to a corresponding relative squared error approach. Two illustrative special cases are discussed, where a generalized least squares estimator on the one hand and a general ridge or Kuks–Olman estimator on the other hand turn out to be uniformly best.     

Enhancing Self‐Protective Behavior: Efficacy Beliefs and Peer Feedback in Risk Communication

In times of a high‐impact safety incident citizens may have a variety of sources available to help them cope with the situation. This research focuses on the interplay of efficacy information in risk communication messages and peer feedback, such as responses on social network sites (SNSs) in the context of a high‐impact risk on the intention to engage in self‐protective behavior. The study pitted high and low efficacy information messages against supporting and opposing peer feedback (N = 242). Results show a significant interaction effect between efficacy information in a news article and peer feedback from SNS messages on both the intention to engage in self‐protective behavior and levels of involvement. Participants who received the article with more efficacy information and also received supportive peer feedback via SNS messages were more likely to express higher levels of involvement and greater intentions to engage in protective behavior. When confronted with a low efficacious news article, the effect of peer feedback on these two variables was significantly stronger. Finally, implications for theory and government risk communication are discussed.     

Informant accuracy in social-network data V. An experimental attempt to predict actual communication from recall data

This paper seeks to discover whether the known inaccuracy of informant recall about their communication behavior can be accounted for by experimentally varying the time period over which recall takes place. The experiment took advantage of a new communications medium (computer conferencing) which enabled us to monitor automatically all the interactions involving a subset of the computer network. The experiment itself was administered entirely by the computer, which interviewed informants and recorded their responses. Variations in time period failed to account for much of the inaccuracy, which continues, as in previous experiments at an unacceptably high level. One positive finding did emerge: although the informants did not know with whom they communicated, the informants en masse seemed to know certain broad facts about the communication pattern. All other findings were negative. For example, it is impossible to predict the people an informant claimed to communicate with but did not; and it is impossible to predict who the five people are that an informant forgot to mention that she or he had communication with. Thus, despite their presumed good intentions, our findings here confirm what we have learned from six previous experiments: What people say about their communications bears no resemblance to their behavior. This suggests that other forms of data gathering, based on questions which require that informants recall their behavior, may well be suspect.     

Competing statistical methods for the fitting of normal species sensitivity distributions: recommendations for practitioners

A species sensitivity distribution (SSD) models data on toxicity of a specific toxicant to species in a defined assemblage. SSDs are typically assumed to be parametric, despite noteworthy criticism, with a standard proposal being the log-normal distribution. Recently, and confusingly, there have emerged different statistical methods in the ecotoxicological risk assessment literature, independent of the distributional assumption, for fitting SSDs to toxicity data with the overall aim of estimating the concentration of the toxicant that is hazardous to % of the biological assemblage (usually with small). We analyze two such estimators derived from simple linear regression applied to the ordered log-transformed toxicity data values and probit transformed rank-based plotting positions. These are compared to the more intuitive and statistically defensible confidence limit-based estimator. We conclude based on a large-scale simulation study that the latter estimator should be used in typical assessments where a pointwise value of the hazardous concentration is required.