Bayesian Information Criterion Statistics Research Articles

Small-area deprivation index does not improve the capability of multisource comorbidity score in mortality prediction.

The stratification of the general population according to health needs allows to provide better-tailored services. A simple score called Multisource Comorbidity Score (MCS) has been developed and validated for predicting several outcomes. The aim of this study was to evaluate whether the ability of MCS in predicting 1-year mortality improves by incorporating socioeconomic data (as measured by a deprivation index). Beneficiaries of the Italian National Health Service who in the index year (2018) were aged 50-85 years and were resident in the Sicily region for at least 2 years were identified. For each individual, the MCS was calculated according to his/her clinical profile, and the deprivation index of the census unit level of the individual's residence was collected. Frailty models were fitted to assess the relationship between the indexes (MCS and deprivation index) and 1-year mortality. Akaike information criterion and Bayesian information criterion statistics were used to compare the goodness of fit of the model that included only MCS and the model that also contained the deprivation index. The models were further compared by means of the area under the receiver operating characteristic curve (AUC). The final cohort included 1,062,221 individuals, with a mortality rate of 15.6 deaths per 1,000 person-years. Both MCS and deprivation index were positively associated with mortality.The goodness of fit statistics of the two models were very similar. For MCS only and MCS plus deprivation index models, Akaike information criterion were 17,013 and 17,038, respectively, whereas Bayesian information criterion were 16,997 and 17,000, respectively. The AUC values were 0.78 for both models. The present study shows that socioeconomic features as measured by the deprivation index did not improve the capability of MCS in predicting 1-year risk of death. Future studies are needed to investigate other sources of data to enhance the risk stratification of populations.

MODELAGEM DA DISTRIBUIÇÃO DIAMÉTRICA DE ESPÉCIES ARBÓREAS EM FRAGMENTO DE FLORESTAS ESTACIONAL SEMIDECIDUAL

ABSTRACT Modeling diameter distribution in natural forests is an important tool for understanding the native woody species dynamics, supporting decision-making for degraded lands management and restoration. Therefore, this work aimed to fit probabilistic density functions to evaluate the diameter structure of three tree species with economic interest, such as Campomansesia xanthocarpa Marl. Ex. O. Berg, Piptadenia gonoacantha (Mart.) J.F. Macbr. and Zeyheria tuberculosa (Vell.) Bureau ex Verl., in a Semideciduous Seasonal Forest fragment at São Paulo State, Brazil. The data came from 83 temporary plots of 10 m x 20 m systematically distributed along 164 ha. Log-normal, Gamma, and Weibull functions were fitted to the three species using the fitdistrplus-package in the R program. Kolmogorov-Smirnov's adherence test was used to evaluate the fits at a 5% probability level. The functions were selected employing Akaike's Information Criterion (AIC) and Schwarz's Bayesian Information Criterion (BIC), in addition to a graphical analysis of the fitted functions. The results indicated that the three species diameter structure is positively asymmetric, representing the exponential pattern, representing continuous natural regeneration. AIC and BIC statistics indicated the Log-normal function to describe the diameter distribution of C. xanthocarpa and Z. tuberculosa, while the Gamma function was the most appropriate for P. gonoacantha. For the three species, the graphical analysis showed the Gamma function results in the best fit without tendency for estimating frequency density per diameter class.

Linear Mixed Model Analysis of Polygenic Hazard Score on Verbal Memory Decline in Alzheimer's Disease.

Alzheimer's disease (AD) is a chronic, progressive, degenerative disease characterized by cognitive dysfunction, including verbal memory loss. Studies were lacking in examining the longitudinal effect of polygenic hazard score on the Rey Auditory Verbal Learning Test-Delayed Total (AVDELTOT) score (a common measure of verbal memory). A key step in analyzing longitudinal changes in cognitive measures using a linear mixed model (LMM) is choosing a suitable covariance structure. The study aims to determine the association between the polygenic hazard score and the AVDELTOT score accounting for repeated measures (the covariance structure). The AVDELTOT scores were collected at baseline, 12 months, 24 months, 36 months, and 48 months from 283 participants with AD, 347 with cognitive normal, and 846 with mild cognitive impairment in the Alzheimer's Disease Neuroimaging Initiative. The Bayesian information criterion statistic was used to select the best covariance structure from 10 covariance structures in longitudinal analysis of AVDELTOT scores. The multivariable LMM was used to investigate the effect of polygenic hazard score status (low vs. medium vs. high) on changes in AVDELTOT scores while adjusted for age, gender, education, APOE-ε4 genotype, and baseline Mini-Mental State Examination score. One-way analysis of variance revealed significant differences in AVDELTOT scores, Mini-Mental State Examination scores, and polygenic hazard scores among AD diagnoses at baseline. Bayesian information criterion favored the compound symmetry covariance structure in the LMM analysis. Using the multivariate LMM, the APOE-ε4 allele and high polygenic hazard score value was significantly associated with AVDELTOT declines. Significant polygenic hazard score status by follow-up visit interactions was discovered. Our findings provide the first evidence of the effect of polygenic hazard score status and APOE-ε4 allele on declines in verbal memory in people with AD.

Prognostic impact of socioeconomic status compared to overall stage for HPV-negative head and neck squamous cell carcinoma

ObjectiveTo estimate the relative prognostic ability of socioeconomic status (SES) compared to overall stage for HPV-negative head and neck squamous cell carcinoma (HNSCC) Materials and methodsData were obtained from the Carolina Head and Neck Cancer Epidemiology Study (CHANCE). An empirical 4-category SES classification system was created. Cox proportional hazards models, survival gradients, Bayesian information criterion (BIC), and Harrell’s C index were used to estimate the prognostic ability of SES compared to stage on overall survival (OS). ResultsThe sample consisted of 1229 patients with HPV-negative HNSCC. Patients with low SES had significantly increased risk of mortality at 5 years compared to patients with high SES (HR 3.11, 95% CI 2.07–4.67; p < 0.001), and the magnitude of effect was similar to overall stage (HR 3.01, 95% CI 2.35–3.86; p < 0.001 for stage IV versus I). Compared to overall stage, the SES classification system had a larger total survival gradient (35.8% vs. 29.1%), similar model fit (BIC statistic of 7412 and 7388, respectively), and similar model discriminatory ability (Harrell’s C index of 0.61 and 0.64, respectively). The association between low SES and OS persisted after adjusting for age, sex, race, alcohol, smoking, overall stage, tumor site, and treatment in a multivariable model (HR 2.96, 95% CI 1.92–4.56; p < 0.001). ConclusionSES may have a similar prognostic ability to overall stage for patients with HPV-negative HNSCC. Future research is warranted to validate these findings and identify evidence-based interventions for addressing barriers to care for patients with HNSCC.

Insight into the longitudinal relationship between chronic subclinical inflammation and obesity from adolescence to early adulthood: a dual trajectory analysis.

This study aimed to understand the longitudinal relationship between C-reactive protein (CRP) and body mass index (BMI) from adolescence to early adulthood. CRP and BMI were collected from participants of the Raine Study Gen2 at 14-, 17-, 20- and 22-year follow-ups (n = 1312). A dual trajectory analysis was conducted to assess the association between CRP and BMI trajectories, providing conditional probabilities of membership of CRP trajectory membership given BMI trajectory membership. Best model fit was assessed by systematically fitting two to eight trajectory groups with linear and quadratic terms and comparing models according to the Bayesian Information Criterion statistic. The three CRP trajectories were; "stable-low" (71.0%), "low-to-high" (13.8%) and "stable-high" (15.2%). Participants in a "high-increasing" BMI trajectory had a higher probability of being in the "stable-high" CRP trajectory (60.4% of participants). In contrast, individuals in the "medium-increasing" BMI trajectory did not have a significantly increased probability of being in the "stable-high" CRP trajectory. These findings support that chronic sub-clinical inflammation is present through adolescence into early adulthood in some individuals. Targeting chronic sub-clinical inflammation though obesity prevention strategies may be important for improving future health outcomes.

Diet scores and prediction of general and abdominal obesity in the Melbourne collaborative cohort study.

To ascertain which of the Alternative Healthy Eating Index (AHEI) 2010, Dietary Inflammatory Index (DII®) and Mediterranean Diet Score (MDS) best predicted BMI and waist-to-hip circumference ratio (WHR). Body size was measured at baseline (1990-1994) and in 2003-2007. Diet was assessed at baseline using a FFQ, along with age, sex, socio-economic status, smoking, alcohol drinking, physical activity and country of birth. Regression coefficients and 95 % CI for the association of baseline dietary scores with follow-up BMI and WHR were generated using multivariable linear regression, adjusting for baseline body size, confounders and energy intake. Population-based cohort in Melbourne, Australia. Included were data from 11 030 men and 16 774 women aged 40-69 years at baseline. Median (IQR) follow-up was 11·6 (10·7-12·8) years. BMI and WHR at follow-up were associated with baseline DII® (Q5 v. Q1 (BMI 0·41, 95 % CI 0·21, 0·61) and WHR 0·009, 95 % CI 0·006, 0·013)) and AHEI (Q5 v. Q1 (BMI -0·51, 95 % CI -0·68, -0·35) and WHR -0·011, 95 % CI -0·013, -0·008)). WHR, but not BMI, at follow-up was associated with baseline MDS (Group 3 most Mediterranean v. G1 (BMI -0·05, 95 % CI -0·23, 0·13) and WHR -0·004, 95 % CI -0·007, -0·001)). Based on Akaike's Information Criterion and Bayesian Information Criterion statistics, AHEI was a stronger predictor of body size than the other diet scores. Poor quality or pro-inflammatory diets predicted overall and central obesity. The AHEI may provide the best way to assess the obesogenic potential of diet.

Development and evaluation of a matrix for assessing fatigue-related risk, derived from a national survey of nurses' work patterns

BackgroundMultiple aspects of nurses’ rosters interact to affect the quality of patient care they can provide and their own health, safety and wellbeing. Objectives(1) Develop and test a matrix incorporating multiple aspects of rosters and recovery sleep that are individually associated with three fatigue-related outcomes - fatigue-related clinical errors, excessive sleepiness and sleepy driving; and (2) evaluate whether the matrix also predicts nurses’ ratings of the effects of rosters on aspects of life outside work. DesignDevelop and test the matrix using data from a national survey of nurses’ fatigue and work patterns in six hospital-based practice areas with high fatigue risk. MethodsSurvey data included demographics, work patterns (previous 14 days), choice about shifts, and the extent to which work patterns cause problems with social life, home life, personal relationships, and other commitments (rated 1 = not at all to 5 = very much). Matrix variables were selected based on univariate associations with the fatigue-related outcomes, limits in the collective employment contract, and previous research. Each variable was categorised as lower (score 0), significant (score 1), or higher risk (score 2). Logistic multiple regression modelling tested the independent predictive power of matrix scores against models including all the (uncategorised) work pattern and recovery sleep variables with significant univariate associations with each outcome variable. Model fit was measured using Akaike and Bayesian Information Criterion statistics. ResultsData were included from 2358 nurses who averaged at least 30 h/week in the previous fortnight in one of the target practice areas. Final matrix variables were: total hours worked; number of shift extensions >30 min, night shifts; breaks < 9 h; breaks ≥ 24 h; nights with sleep 11pm to 7am; days waking fully rested; and roster change.After controlling for gender, ethnicity, years of nursing experience, and the extent of shift choice, the matrix score was a significant independent predictor of each of the three fatigue-related outcomes, and for all four aspects of life outside work. For all outcome variables, the model including the matrix score was a better fit to the data than the equivalent model including all the (uncategorised) work pattern variables. ConclusionsA matrix that predicts the likelihood of nurses reporting fatigue-related safety outcomes can be used to compare the impact of rosters both at work and outside work. It can be used for roster design and management, and to guide nurses’ choices about the shifts they work.

The role of maternal homocysteine concentration in placenta-mediated complications: findings from the Ottawa and Kingston birth cohort

BackgroundHomocysteine is an intermediate metabolite implicated in the risk of placenta-mediated complications, including preeclampsia, placental abruption, fetal growth restriction, and pregnancy loss. Large cohort and case-control studies have reported inconsistent associations between homocysteine and these complications. The purpose of this study was to investigate whether elevated maternal plasma homocysteine concentration in the early to mid-second trimester is associated with an increased risk of placenta-mediated complications. We examined the following potential moderating factors that may explain discrepancies among previous studies: high-risk pregnancy and the MTHFR 677C>T polymorphism.MethodsWe analyzed data from participants recruited to the Ottawa and Kingston (OaK) Birth Cohort from 2002 to 2009 in Ottawa and Kingston, Canada. The primary outcome was a composite of any placenta-mediated complication, defined as a composite of small for gestational age (SGA) infant, preeclampsia, placental abruption, and pregnancy loss. Secondary outcomes were, individually: SGA infant, preeclampsia, placental abruption, and pregnancy loss. We conducted multivariable logistic regression analyses with homocysteine as the primary continuous exposure, adjusting for gestational age at the time of bloodwork and explanatory maternal characteristics. The functional form, i.e., the shape of the homocysteine association with the outcome was examined using restricted cubic splines and information criteria (Akaike’s/Bayesian Information Criterion statistics). Missing data were handled with multiple imputation.Results7587 cohort participants were included in the study. Maternal plasma homocysteine concentration was significantly associated (linearly) with an increased risk of both the composite outcome of any placenta-mediated complication (p = 0.0007), SGA (p = 0.0010), severe SGA, and marginally with severe preeclampsia, but not preeclampsia, placental abruption and pregnancy loss. An increase in homocysteine concentration significantly increased the odds of any placenta-mediated complication (odds ratio (OR) for a 5 μmol/L increase: 1.63, 95% Confidence Interval (CI) 1.23–2.16) and SGA (OR 1.76, 95% CI 1.25–2.46). Subgroup analyses indicated some potential for modifying effects of the MTHFR 677C>T genotype and high-risk pregnancy, although the interaction was not statistically significant (high-risk subgroup OR 2.37, 95% CI 1.24–4.53, p-value for interaction =0.14).ConclusionsOur results suggest an independent effect of early to mid-pregnancy elevated maternal homocysteine on placenta-mediated pregnancy complications.

Establishment of an early warning system for cutaneous leishmaniasis in Fars province, Iran

Objective: To establish an early warning system for cutaneous leishmaniasis in Fars province, Iran in 2016. Methods: Time-series data were recorded from 29 201 cutaneous leishmaniasis cases in 25 cities of Fars province from 2010 to 2015 and were used to fit and predict the cases using time-series models. Different models were compared via Akaike information criterion/ Bayesian information criterion statistics, residual analysis, autocorrelation function, and partial autocorrelation function sample/model. To decide on an outbreak, four endemic scores were evaluated including mean, median, mean + 2 standard deviations, and median + interquartile range of the past five years. Patients whose symptoms of cutaneous leishmaniasis began from 1 January 2010 to 31 December 2015 were included, and there were no exclusion criteria. Results: Regarding four statistically significant endemic values, four different cutaneous leishmaniasis space-time outbreaks were detected in 2016. The accuracy of all four endemic values was statistically significant (P Conclusions: This study presents a protocol to set early warning systems regarding time and space features of cutaneous leishmaniasis in four steps: (i) to define endemic values based on which we could verify if there is an outbreak, (ii) to set different time-series models to forecast cutaneous leishmaniasis in future, (iii) to compare the forecasts with endemic values and decide on space-time outbreaks, and (iv) to set an alarm to health managers.

Analysis of definitive screening designs: Screening vs prediction

The use of definitive screening designs (DSDs) has been increasing since their introduction in 2011. These designs are used to screen factors and to make predictions. We assert that the choice of analysis method for these designs depends on the goal of the experiment, screening, or prediction. In this work, we present simulation results to address the explanatory (screening) use and the predictive use of DSDs. To address the predictive ability of DSDs, we use two 5‐factor DSDs and simultaneously run central composite designs case studies on which we will compare several common analysis methods. Overall, we find that for screening purposes, the Dantzig selector using the Bayesian Information Criterion statistic is a good analysis choice; however, when the goal of analysis is prediction forward selection using the Bayesian Information Criterion statistic produces models with a lower mean squared prediction error.

Issues Encountered When Modelling The Long Term Clinical Effectiveness Of Treatments From Short Term Trials Using Second-Line Treatments For Advanced Or Metastatic Renal Cell Carcinoma (AMRCC) As An Example

Long term effectiveness data on progression-free survival (PFS) and overall survival (OS) are rarely available from efficacy trials. Methods to estimate the expected PFS and OS using a range of survival models were explored up to 30 years for patients receiving axitinib, cabozantinib, everolimus, nivolumab, and BSC for second-line amRCC. Several parametric survival models, including cubic splines, were fitted to the everolimus and nivolumab groups of the CheckMate 025 trial to provide baseline curves for PFS and OS. Model fit was assessed using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), as well as clinical expert opinion to assess the plausibility of extrapolations. Hazard ratios (HRs), derived from a mixed treatment comparison (MTC), were applied to the everolimus curves to produce estimates for cabozantinib and BSC. A reliable HR could not be derived for axitinib due to a violation of proportional hazards (PH), as well as patient cross-over in the TARGET trial. Therefore, it was assumed that axitinib was equivalent to everolimus. For OS, there were slight differences in the AIC and BIC statistics across the treatment groups, but the best fitting model was considered to be the Weibull. For PFS, the best fitting model was the 2-knot spline, clearly indicated by the lowest AIC and BIC for both everolimus and nivolumab. The Weibull models produced a mean OS of 26.5, 38.1, 26.5, 30.3 and 15.0 months, for axitinib, cabozantinib, everolimus, nivolumab and BSC, respectively. The equivalent values for the 2-knot spline PFS models were 7.9, 15.9, 7.9, 10.8 and 2.8 months, respectively. The assumption of PH can be a limitation when comparing multiple treatments across different trials using an MTC. Practical solutions such as choosing the appropriate “baseline” trial and assuming clinical equivalence, where plausible, can help mitigate concerns in some cases.

Comparisons of Five Performance Validity Indices in Bona Fide and Simulated Traumatic Brain Injury

A number of performance validity tests (PVTs) are used to assess memory complaints associated with traumatic brain injury (TBI); however, few studies examine the concordance and predictive accuracy of multiple PVTs, specifically in the context of combined models in known-group designs. The present study compared five widely used PVTs: the Test of Memory Malingering (TOMM), Medical Symptom Validity Test (MSVT), Reliable Digit Span (RDS), Word Choice Test (WCT), and California Verbal Learning Test – Forced Choice (CVLT-FC). Participants were 51 adults with bona fide moderate to severe TBI and 58 demographically comparable healthy adults coached to simulate memory impairment. Classification accuracy of individual PVTs was evaluated using logistic regression and receiver operating characteristic (ROC) curves, examining both the dichotomous cutting scores as recommended by the test publishers and continuous scores for the measures. Results demonstrated nearly equivalent discrimination ability of the TOMM, MSVT, and CVLT-FC as individual predictors, all of which markedly outperformed the WCT and RDS. Models of combined PVTs were examined using Bayesian information criterion statistics, with results demonstrating that diagnostic accuracy showed only small to modest growth when the number of tests was increased beyond two. Considering the clinical and pragmatic issues in deriving a parsimonious assessment battery, these findings suggest that using the TOMM and CVLT in conjunction or the MSVT and CVLT in conjunction maximized predictive accuracy as compared to a single index or an assortment of these widely used measures.

A latent class distance association model for cross‐classified data with a categorical response variable

In this paper we propose a latent class distance association model for clustering in the predictor space of large contingency tables with a categorical response variable. The rows of such a table are characterized as profiles of a set of explanatory variables, while the columns represent a single outcome variable. In many cases such tables are sparse, with many zero entries, which makes traditional models problematic. By clustering the row profiles into a few specific classes and representing these together with the categories of the response variable in a low-dimensional Euclidean space using a distance association model, a parsimonious prediction model can be obtained. A generalized EM algorithm is proposed to estimate the model parameters and the adjusted Bayesian information criterion statistic is employed to test the number of mixture components and the dimensionality of the representation. An empirical example highlighting the advantages of the new approach and comparing it with traditional approaches is presented.

Adaptive Learning in Bayesian Networks for Incident Duration Prediction

The development of a practical model for incident management is investigated through Bayesian networks (BNs) in this study. BNs are capable of accurately predicting incident durations and can easily be incorporated into incident management activities of traffic management centers to improve the real-time decision-making process. Three structure learning algorithms were used to construct BN structures. They were estimated by using 2005 New Jersey incident data; the best-performing one was chosen for the incident duration prediction with the use of the 10-fold cross-validation method and the Bayesian information criterion statistic. To demonstrate the performance of Bayesian learning, the chosen model was fed by 2011 New Jersey incident data on a monthly and quarterly basis. Comparing the prediction results for 2011 data with and without adaptive learning showed that the developed BN had the capability to automatically adapt itself to future conditions by learning the patterns of new incidents and their respective durations.

The Trauma and Injury Severity Score (TRISS) revised

BackgroundThe Trauma and Injury Severity Score (TRISS) remains the most commonly used tool for benchmarking trauma fatality outcome. Recently, it was demonstrated that the predictive power of TRISS could be substantially improved by re-classifying the component variables and treating the variable categories nominally. This study aims to develop revised TRISS models using re-classified variables, to assess these models’ predictive performances against existing TRISS models, and to identify and recommend a preferred TRISS model. Materials and methodsRevised TRISS models for blunt and penetrating injury mechanism were developed on an adult (aged ≥15 years) sample from the National Trauma Data Bank National Sample Project (NSP), using 5-category variable classifications and weighted logistic regression. Their predictive performances were then assessed against existing TRISS models on the unweighted NSP, National Trauma Data Bank (NTDB), and New Zealand Database (NZDB) samples using area under the Receiver Operating Characteristic curve (AUC) and Bayesian Information Criterion (BIC) statistics. ResultsThe weighted NSP sample included 1,124,001 adults with blunt or penetrating injury mechanism events and known discharge status, of whom 1,061,709 (94.5%) survived to discharge. Complete information for all TRISS variables was available for 896,212 (79.7%). Revised TRISS models that included main-effects and two-factor interaction terms had superior AUC and BIC statistics to main-effects models and existing TRISS models for patients with complete data in NSP, NTDB and NZDB samples. Predictive performance decreased as the number of variables with missing values included within revised TRISS models increased, but model performances generally remained superior to existing TRISS models. DiscussionRevised TRISS models had importantly improved predictive capacities over existing TRISS models. Additionally, they were easily computed, utilised only those variables already collected for existing TRISS models, and could be applied and produce meaningful survival probabilities when one or more of the predictor variables contained missing values. The preferred revised TRISS model included main-effects and two-factor interaction terms and allowed for missing values in all predictor variables. A strong case exists for replacing existing TRISS models in trauma scoring systems benchmarking software with this preferred revised TRISS model.

Trauma and Injury Severity Score (TRISS): Is it time for variable re-categorisations and re-characterisations?

BackgroundDespite its limitations, the Trauma and Injury Severity Score (TRISS) continues to be the most commonly used tool for benchmarking trauma outcome. Since its inception, considerable energy has been devoted to improving TRISS. However, there has been no investigation into the classification or characterisation of the TRISS variables. Using a major nationally representative database, this study aims to explore the adequacy of the existing TRISS model by investigating variable re-categorisations and alternative characterisations in a logistic model used to predict survival in adults after traumatic injury. Materials and methodsData were obtained from the National Trauma Data Bank National Sample Project (NSP). Each variable in the TRISS model was related to discharge status and various categorisations considered using weighted logistic regression. Categorisations were treated nominally, using a series of indicator variables. For each variable and classification level, the best category combination was ascertained using the Bayesian Information Criterion (BIC). All best 5-category classified TRISS variables were combined, as were all best 10-category classified TRISS variables, and their predictive performance assessed against two conventionally defined TRISS models on the unweighted NSP sample using area under the Receiver Operating Characteristic curve (AUC) and BIC statistics. ResultsOverall, the weighted sample included 1,124,001 adults with injury events and known discharge status, of whom 1,061,709 (94.5%) were alive at discharge. When separately related to discharge status, each re-classified TRISS variable yielded a superior BIC statistic to its original specification. When investigating predictive performance, complete information was available for 167,239 (79.9%) adults with blunt and 20,643 (82.3%) adults with penetrating injury mechanisms. AUC and BIC estimates for the re-classified TRISS models were superior to the conventionally defined TRISS models. While having better predictive precision, the complexity associated with the best 10-category model resulted in the best 5-category model being preferred for penetrating mechanism injuries and being negligibly inferior for blunt mechanism injuries. DiscussionSubstantial improvements in the predictive power of TRISS were demonstrated by re-classifying the component variables and treating the variable categories nominally. However, before a new TRISS model with updated coefficients can be published, variable interactions and the effect of missing data needs thorough statistical evaluation.

Finding Jumps in Otherwise Smooth Curves: Identifying Critical Events in Political Processes

Many social processes are stable and smooth in general, with discrete jumps. We develop a sequential segmentation spline method that can identify both the location and the number of discontinuities in a series of observations with a time component, while fitting a smooth spline between jumps, using a modified Bayesian Information Criterion statistic as a stopping rule. We explore the method in a large-n, unbalanced panel setting with George W. Bush's approval data, a small-n time series with median DW-NOMINATE scores for each Congress over time, and a series of simulations. We compare the method to several extant smoothers, and the method performs favorably in terms of visual inspection, residual properties, and event detection. Finally, we discuss extensions of the method.

Collision Models for Multilane Highway Segments to Examine Safety of Curbs

The objective of this study was to develop valid statistical collision models for multilane highway segments to examine the safety of curbs. For this, road geometric traffic and collision data for 2001 to 2003 were collected. The data included 2,274 collisions and 885 injury collisions that occurred on 191.85 mi of 199 directional segments in North Carolina. The authors applied a new modeling method of introducing variables into the model one by one in a multiplicative form. A nonlinear optimizing algorithm for estimating parameters using a negative binomial log likelihood function was adopted for the modeling. The functional form of the variable to be introduced was determined on the basis of the relationship between the recorded number of collisions and the number of collisions predicted by the current model without the variable. The integrate-differentiate method was applied to find candidate functional forms for each variable. Model selections were based on the −2 log likelihood and Bayesian information criterion statistics, and the cumulative residuals plot method to check the goodness of fit of the models was adopted. As a result of the modeling efforts, the annual average daily traffic, access point density, shoulder width, and shoulder type (including curb presence) variables were introduced to the final model for total collisions. The same variables except the shoulder type variable were introduced to the injury collision model. Overall, then, it appears that curbs mean fewer total collisions and no change in injury collisions as compared to no curbs on the sampled road segments.

A simplified lesion classification for predicting success and complications of coronary angioplasty

In 1988, the American College of Cardiology/American Heart Association (ACC/AHA) Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures presented a classification of coronary lesions utilizing 26 lesion features to predict the success and complications of balloon angioplasty. Using data from the Registry of the Society for Cardiac Angiography and Interventions (SCAI) we evaluated the ability of this classification to predict success and complications. Lesion success, death in hospital, emergency cardiac bypass surgery, and major adverse events were evaluated in 41,071 patients who underwent single-vessel angioplasty from January 1993 to June 1996. Logistic models using the ACC/AHA lesion classification, vessel patency, or both, were compared. A new classification based on the interaction of the ACC/AHA classification plus lesion patency was compared with the existing ACC/AHA classification. Vessel patency, added to the ACC/AHA classification, improved prediction of lesion success (p ≤0.0001). Class A and patent B lesions had similar success and complication rates, so a simplified classification (SCAI) using only 7 lesion characteristics could be created. This system (I: non-C patent, II: C patent, III: non-C occluded, and IV: C occluded) improved prediction of lesion success compared with the ACC/AHA classification (Bayesian Information Criterion statistic: ACC/AHA 16539, SCAI 15956; and area under the receiver- operating characteristics curve 0.659, 0.693, respectively). The SCAI classification was preferred for predicting major complications and in-hospital death and was similar to the ACC/AHA classification for predicting emergency bypass surgery.