Lynne Seymour

An update of U.S. temperature trends

Abstract This note updates the temperature trend study of the contiguous 48 United States by Lund et al. for data observed during the 4-yr period 1997–2000. A parsimonious changepoint parameterization is now used, and the methods for handling missing data are improved. The number of stations with useable data has now increased from 359 to 969, thereby improving the accuracy of the reported spatial patterns in the trends. The average record length of the 969 stations is now 103 yr, with the longest record starting in 1812 and the shortest in 1926. The methodological improvements and additional 4 yr of data produce slightly smaller trend estimate standard errors. Warming is found in the Northeast, West, and northern Midwest, with slight cooling in the Southeast; overall, the trends here suggest more warming than those of the Lund et al. study.

Characterization of Three-Dimensional Spatial Aggregation and Association Patterns of Brown Rot Symptoms within Intensively Mapped Sour Cherry Trees

BACKGROUND AND AIMS Characterization of spatial patterns of plant disease can provide insights into important epidemiological processes such as sources of inoculum, mechanisms of dissemination, and reproductive strategies of the pathogen population. Whilst two-dimensional patterns of disease (among plants within fields) have been studied extensively, there is limited information on three-dimensional patterns within individual plant canopies. Reported here are the detailed mapping of different symptom types of brown rot (caused by Monilinia laxa) in individual sour cherry tree (Prunus cerasus) canopies, and the application of spatial statistics to the resulting data points to determine patterns of symptom aggregation and association. METHODS A magnetic digitizer was utilized to create detailed three-dimensional maps of three symptom types (blossom blight, shoot blight and twig canker) in eight sour cherry tree canopies during the green fruit stage of development. The resulting point patterns were analysed for aggregation (within a given symptom type) and pairwise association (between symptom types) using a three-dimensional extension of nearest-neighbour analysis. KEY RESULTS Symptoms of M. laxa infection were generally aggregated within the canopy volume, but there was no consistent pattern for one symptom type to be more or less aggregated than the other. Analysis of spatial association among symptom types indicated that previous years twig cankers may play an important role in influencing the spatial pattern of current years symptoms. This observation provides quantitative support for the epidemiological role of twig cankers as sources of primary inoculum within the tree. CONCLUSIONS Presented here is a new approach to quantify spatial patterns of plant disease in complex fruit tree canopies using point pattern analysis. This work provides a framework for quantitative analysis of three-dimensional spatial patterns within the finite tree canopy, applicable to many fields of research.

Assessment of Changes in Utilization of Health-Care Services after Implementation of a Prior Authorization Policy for Atypical Antipsychotic Agents

BACKGROUND: In 2004, the Georgia Medicaid program implemented a prior authorization (PA) policy for certain atypical antipsychotic agents, resulting in an average savings of

Methods for Starting the Detection of Undocumented Multiple Changepoints

2.7 million per year. OBJECTIVE: To determine whether implementation of a PA policy for atypical antipsychotic drugs increased health-care utilization in the Georgia Medicaid program from July 2003 to April 2006. METHODS: A single cohort observational study employing segmented regression and time series analysis was conducted to determine health-care services utilization, including emergency department (ED) visits, outpatient office visits, hospital admissions, and length of stay (LOS). Study subjects included continuously eligible adult Georgia Medicaid recipients with a schizophrenia-related diagnosis and documented use of an atypical antipsychotic medication (N = 12,120). Where applicable, analysis of a noncontinuously eligible population was also performed to investigate disenrollment bias in study results. RESULTS: A significant decline in post-policy trend for the average number of ED visits (absolute difference −0.042 per member per month (PMPM); relative difference −20.92%) and average number of hospital admissions PMPM (absolute difference −0.010 PMPM; relative difference −22.27%) was observed at the end of the study period. Baseline and pre-policy trends were found to be significant predictors for both endpoints. Significant models were not identified for average outpatient office visits PMPM or average LOS per admission. CONCLUSIONS: In contrast to other published studies on PA for atypical antipsychotics, patient outcomes improved after the initiation of the policy. To the extent that medical utilization reflects patient health outcomes and health status, the results of this study indicate that the PA program has potentially improved the health of schizophrenic patients in Georgia and lowered program costs.

Spatio-temporal Patterns of Pre-harvest Brown Rot Epidemics within Individual Peach Tree Canopies

Abstract Four algorithms are given, as a first step toward the practical detection of undocumented multiple changepoints. These algorithms are based on the two-phase regression method of Lund and Reeves, as well as the robust method of Lazante. The result of each method is a set that contains statistically detectable changepoints; each candidate is then either independently validated as a changepoint or discarded. This is demonstrated and the methods are compared on artificial data, and then the methods are implemented on streamflow data from the Flint River in southwest Georgia. Most notably, the method based on two-phase regression was able to detect a well-known yet undocumented drop in streamflow from a local drought that no other methods have so far been able to detect.

Approximate bayes model selection procedures for Gibbs-Markov random fields

Tree canopies are architecturally complex and pose several challenges for measuring and characterizing spatial patterns of disease. Recently developed methods for fine-scale canopy mapping and three-dimensional spatial pattern analysis were applied in a 3-year study to characterize spatio-temporal development of pre-harvest brown rot of peach, caused by Monilinia fructicola, in 13 trees of different maturity classes. We observed a negative correlation between an index of disease aggregation and disease incidence in the same tree (r = −0.653, P < 0.0001), showing that trees with higher brown rot incidence had lower aggregation of affected fruit in their canopies. Significant (P ≤ 0.05) within-canopy aggregation among symptomatic fruit was most pronounced for early-maturing cultivars and/or early in the epidemic. This is consistent with the notion of a greater importance of localized, within-tree sources of inoculum at the beginning of the epidemic. Four of five trees having >10 blossom blight symptoms per tree showed a significant positive spatial association of pre-harvest fruit rot to blossom blight within the same canopy. Spatial association analyses further revealed one of two outcomes for the association of new fruit rot symptoms with previous fruit rot symptoms in the same tree, whereby the relationship was either not significant or exhibited a significant negative association. In the latter scenario, the newly diseased fruit were farther apart from previously symptomatic fruit than expected by random chance. This unexpected result could have been due to uneven fruit ripening in different sectors of the canopy, which could have affected the timing of symptom development and thus led to negative spatial associations among symptoms developing over time in a tree.

Assessing temperature anomalies for a geographical region : A control chart approach

For applications in texture synthesis, we derive two approximate Bayes criteria for selecting a model from a collection of Markov random fields. The first criterion is based on a penalized maximum likelihood. The second criterion, a Markov chain Monte Carlo approximation to the first, has distinct computational advantages. Some simulation results are also presented.

Modeling plant growth using regression with periodically correlated errors

An X-bar control chart is developed to quantify monthly temperature anomalies for a geographical region. The developed chart accounts for space–time correlations in the temperatures and allows for accurate statistical quantification of temperature fluctuations. The methods presented can also be used to probabilistically quantify the severity of hot and cold periods. Using time series prediction techniques, a control chart for monthly average temperature data is developed for an arbitrary number of recording stations. The proposed chart takes into account trends, seasonality in the means and variances of the temperatures, and perhaps most importantly, the temporal and spatial correlations also present in such data. The results are applied in the analysis of six monthly temperature series from the southeastern United States during 1900–1993. Copyright

Rates of Convergence and Approximations to the Distribution of the Maximum of Chain-Dependent Sequences

Recent developments in the analysis of periodically correlated time series are used to analyze data in a new approach to modeling plant growth. The data—carbon dioxide (CO2) exchange rates in plants—were gathered using a novel and innovative system for near-continuous measurement of CO2 exchange. Least squares estimates are derived for a piecewise polynomial trend, which carries over in an obvious fashion to a general linear model. The residuals from fitting this trend have seasonal components that are well modeled using a periodic autoregressive moving average (PARMA) model. Standard errors for the trend estimates are then derived based on the fitted PARMA model.

Estimating the Dimensionality of a Ridge in Quadratic Response Surface Models

In this paper we propose two methods for approximating the distribution of the maximum for a chain-dependent process. For these methods we investigate the rate of convergence of the approximations. A simulation study was performed that demonstrated excellent performance of the approximations even for small sample sizes. In addition, an application was made to river flow data collected on the North Pacolet River in South Carolina.