Applications of sequential set partitioning: a set of technical information systems problems

Many problems that appear in different contexts are conceptually similar and so are amenable to solution by a common technique. Three such technical Information Systems (IS) problems are: (1) segmentation of computer programs; (2) attribute discretization for decision tree induction; and (3) design of hash tables in database systems. In this paper we show how each of these seemingly different problems can be formulated as a sequential (set) partitioning problem, and solved using a parametric linear programming (LP) procedure. This approach provides optimal solutions unlike previous solution approaches which were either greedy heuristics or limited to solving only a specific problem situation. Given the likelihood that other applications of the sequential partitioning problem exist in IS, the material presented here could be useful to other researchers in formulating the problem at an appropriate level of abstraction so that available optimal solution approaches can be identified. In addition to providing a common solution method, parametric LP frees the user from having to make premature decisions regarding the number of groups for the partition, and this decision can be delayed post solution.     

A Comparison of Three Methods for Calculating Confidence Intervals for the Benchmark Dose

Various methods exist to calculate confidence intervals for the benchmark dose in risk analysis. This study compares the performance of three such methods in fitting nonlinear dose-response models: the delta method, the likelihood-ratio method, and the bootstrap method. A data set from a developmental toxicity test with continuous, ordinal, and quantal dose-response data is used for the comparison of these methods. Nonlinear dose-response models, with various shapes, were fitted to these data. The results indicate that a few thousand runs are generally needed to get stable confidence limits when using the bootstrap method. Further, the bootstrap and the likelihood-ratio method were found to give fairly similar results. The delta method, however, resulted in some cases in different (usually narrower) intervals, and appears unreliable for nonlinear dose-response models. Since the bootstrap method is more time consuming than the likelihood-ratio method, the latter is more attractive for routine dose-response analysis. In the context of a probabilistic risk assessment the bootstrap method has the advantage that it directly links to Monte Carlo analysis.     

A method for estimating age-specific reference intervals ('normal ranges') based on fractional polynomials and exponential transformation

The age-specific reference interval is an important screening tool in medicine. Put crudely, an individual whose value of a variable of interest lies outside certain extreme centiles may be suspected of abnormality. We propose a parametric method for constructing such intervals. It provides smooth centile curves and explicit formulae for the centile estimates and for standard deviation (SD) scores (age-standardized values). Each parameter of an exponential–normal or modulus–exponential–normal density is modelled as a fractional polynomial function of age. Estimation is by maximum likelihood. These three- and four-parameter models involve transformations of the data towards normality which remove non-normal skewness and/or kurtosis. Fractional polynomials provide more flexible curve shapes than do conventional polynomials. The method easily accommodates binary covariates facilitating, for example, parsimonious modelling of age- and sex-specific centile curves. A method of calculating precision profiles for centile estimates is proposed. Goodness of fit is assessed by using Q–Q -plots and Shapiro–Wilk W -tests of the SD scores, and likelihood ratio tests of the parameters of an enlarged model. Four substantial real data sets are used to illustrate the method. Comparisons are made with the semiparametric LMS method of Cole and Green.     

Bootstrap-based evaluation of markov-switching time series models

This paper explores the possibility of evaluating the adequacy of Markov-switching time series models by comparing selected functionals (such as the spectral density function and moving empirical moments) obtained from the data with those of the fitted model using a bootstrap algorithm. The proposed model checking procedure is easy to implement and flexible enough to be adapted to a wide variety of models with parameters subject to Markov regime-switching. Examples with real and artificial data illustrate the potential of the methodology.     

Bivariate Weibull regression model based on censored samples

The most natural parametric distribution to consider is the Weibull model because it allows for both the proportional hazard model and accelerated failure time model. In this paper, we propose a new bivariate Weibull regression model based on censored samples with common covariates. There are some interesting biometrical applications which motivate to study bivariate Weibull regression model in this particular situation. We obtain maximum likelihood estimators for the parameters in the model and test the significance of the regression parameters in the model. We present a simulation study based on 1000 samples and also obtain the power of the test statistics.     

Permutational tests for correlation matrices

Permutational tests are proposed for the hypotheses that two population correlation matrices have common eigenvectors, and that two population correlation matrices are equal. The only assumption made in these tests is that the distributional form is the same in the two populations; they should be useful as a prelude either to tests of mean differences in grouped standardised data or to principal component investigation of such data.The performance of the permutational tests is subjected to Monte Carlo investigation, and a comparison is made with the performance of the likelihood-ratio test for equality of covariance matrices applied to standardised data. Bootstrapping is considered as an alternative to permutation, but no particular advantages are found for it. The various tests are applied to several data sets.     

Properties of Model-Averaged BMDLs: A Study of Model Averaging in Dichotomous Response Risk Estimation

Model averaging (MA) has been proposed as a method of accounting for model uncertainty in benchmark dose (BMD) estimation. The technique has been used to average BMD dose estimates derived from dichotomous dose-response experiments, microbial dose-response experiments, as well as observational epidemiological studies. While MA is a promising tool for the risk assessor, a previous study suggested that the simple strategy of averaging individual models' BMD lower limits did not yield interval estimators that met nominal coverage levels in certain situations, and this performance was very sensitive to the underlying model space chosen. We present a different, more computationally intensive, approach in which the BMD is estimated using the average dose-response model and the corresponding benchmark dose lower bound (BMDL) is computed by bootstrapping. This method is illustrated with TiO(2) dose-response rat lung cancer data, and then systematically studied through an extensive Monte Carlo simulation. The results of this study suggest that the MA-BMD, estimated using this technique, performs better, in terms of bias and coverage, than the previous MA methodology. Further, the MA-BMDL achieves nominal coverage in most cases, and is superior to picking the "best fitting model" when estimating the benchmark dose. Although these results show utility of MA for benchmark dose risk estimation, they continue to highlight the importance of choosing an adequate model space as well as proper model fit diagnostics.     

Comments on “Canonical Correlation in Multivariate Time Series Analysis With an Application to One-Year-Ahead and Multiyear- Ahead Macroeconomic Forecasting,” by P. W. Otter

We examine the orthogonality assumption of seasonal and nonseasonal components for official quarterly unemployment figures in Germany and the United States. Although nonperiodic correlations do not seem to reject the orthogonality assumption, a periodic analysis based on correlation functions that vary with the seasons indicates the violation of orthogonality. We find that the unadjusted data can be described by periodic autoregressive models with a unit root. In simulations we replicate the empirical findings for the German data, where we use these simple models to generate artificial samples. Multiplicative seasonal adjustment leads to large periodic correlations. Additive adjustment leads to smaller ones.     

Finite Sample Comparison of Parametric,Semiparametric, and Wavelet Estimators of Fractional Integration

ABSTRACT

In this paper we compare through Monte Carlo simulations the finite sample properties of estimators of the fractional differencing parameter, d. This involves frequency domain, time domain, and wavelet based approaches, and we consider both parametric and semiparametric estimation methods. The estimators are briefly introduced and compared, and the criteria adopted for measuring finite sample performance are bias and root mean squared error. Most importantly, the simulations reveal that (1) the frequency domain maximum likelihood procedure is superior to the time domain parametric methods, (2) all the estimators are fairly robust to conditionally heteroscedastic errors, (3) the local polynomial Whittle and bias-reduced log-periodogram regression estimators are shown to be more robust to short-run dynamics than other semiparametric (frequency domain and wavelet) estimators and in some cases even outperform the time domain parametric methods, and (4) without sufficient trimming of scales the wavelet-based estimators are heavily biased.     

Decriminalization and Marijuana Smoking Prevalence: Evidence From Australia

This article uses the 2001 National Drug Strategy Household Survey to assess the impact of marijuana decriminalization policy on marijuana smoking prevalence in Australia. Both parametric and nonparametric methods are used. The parametric approach includes endogenous probit switching, two-part, sample selection, and standard dummy variable models, while the nonparametric approach uses propensity score stratification matching. Specification analyses are also conducted. A nonparametric kernel-based test is constructed to select between parametric and nonparametric models, and the likelihood ratio test is used to choose among parametric models. Our analyses favor the endogenous switching model where decriminalization increases the probability of smoking by 16.2%.