Search

PDF file - 856K, Supplementary Table 1, presenting Conditional logistic regression model relating the three biomarkers OGG1, MPG and APE1 to lung cancer. Supplementary Figure 1, presenting Association of DNA repair activity with lung cancer risk, including relative frequency plots for APE1 activity in 100 case patients and 99 matched controls subjects, relative frequency plots for the integrated DNA repair OMA score in 100 case patients and 99 matched controls subjects, and odds ratio calculated as a function of the percentiles of the distribution of the OMA score among the control subjects.

The COVID-19 pandemic cast a dramatic spotlight on the use of data as a fundamental component of good decision-making. Evaluating and comparing alternative policies required information on concurrent infection rates and insightful analysis to project them into the future. Statisticians in Israel were involved in these processes early in the pandemic in some silos as an ad-hoc unorganized effort. Informal discussions within the statistical community culminated in a roundtable, organized by three past presidents of the Israel Statistical Association, and hosted by the Samuel Neaman Institute in April 2021. The meeting was designed to provide a forum for exchange of views on the profession’s role during the COVID-19 pandemic, and more generally, on its influence in promoting evidence-based public policy. This paper builds on the insights and discussions that emerged during the roundtable meeting and presents a general framework, with recommendations, for involving statisticians and statistics in decision-making.

The objectives of this note are to correct a common error and to clarify the connection between the Gini terminology as used in the economic literature and the one used in the diagnostic and classification literature. More specifically, the connection between the area under the receiver operating characteristic (ROC) curve, which is frequently used in the diagnosis and classification literature, and the Gini terminology, which is mainly used in the economic literature, is clarified. It is shown that the area under the ROC curve is related to the covariance between the two vectors $$Y=\{y_i\}_{i=1}^{n_0}$$ and $$\{i/{n_0}\}_{i=1}^{n_0}$$. Here $$y_i$$ is the number of items classified to group 1 lying between the $$(i-1)\mathrm{th}$$ and the $$i\mathrm{th}$$ items classified to group 0, and $$n_0$$ is the number of items in group 0.

Collecting driving speed data is a crucial element in road safety research. Recent technological developments enable a wide range of alternative traffic monitoring possibilities. In particular, the availability of cellular and GPS devices creates the opportunity to utilize probe vehicle data. To date, most studies using probe vehicle data focused on operational purposes such as congestion identification, relying primarily on average speeds. Probe vehicle data can also be useful for safety studies. The data is often collected continuously (24/7, 365 days a year). It can therefore enable detailed analyses of the continuous changes in speed over time. In addition, the spatial coverage of the data enables continuous evaluation of the halo effect of speed enforcement installation, compared with a relatively small number of measurement sites (e.g. 3–5) in traditional methods. Probe vehicle data offer one more advantage: When performing a before-after study, the before measurements have to be performed before camera installation. However, installation plans may change, and therefore conducting the before measurements may be administratively challenging. Probe vehicle data, especially if collected continuously, regardless of research needs, enable researchers to choose the desired sites and times for analysis in retrospect, after the cameras have been installed. The purpose of this paper is to describe the results obtained from analysis of probe vehicle data on the effect of speed enforcement cameras that were installed at non-urban roads in Israel during 2011–2013. In addition to the contribution regarding the impact of enforcement cameras, the paper demonstrates the potential (as well as the challenges) of using probe vehicle data for safety studies. The analyses in this paper consider 22 installation sites and 12 comparison sites. The results show that at the installation sites (within 100 m), the 85th percentile speeds exhibited a short-term (two months) reduction of 2.92 km/h (SD = 2.37) and a long-term (one year) reduction of 6.48 km/h (SD = 3.5). An a-symmetric double exponential model is presented to depict the halo effect up to 1.5 km from the installation location.

Background Nevus phenotype is a marker of melanoma risk. In vivo prediction of microscopic pattern is needed to more precisely classify nevi. Objective To identify dermoscopic and clinical predictors of microscopic patterns of typical nevi. Methods We used reflectance confocal microscopy (RCM) to classify microscopic patterns of nevi. We prospectively accrued adults presenting for periodic skin screening and imaged, with dermoscopy and RCM, 3 randomly selected nevi from the upper and lower back and lower extremity. RCM patterns were classified into ring, clod, meshwork, and composite types. Logistic regression was used to identify best predictors of RCM pattern. Results The study included 310 nevi from 112 participants (mean age 44 years; 51 women). Dermoscopic reticular pattern correlated most frequently (59.9%) with RCM ring pattern, dermoscopic globular with RCM composite (56.6%) and RCM clod (35.9%), dermoscopic complex with RCM composite (76.3%), and dermoscopic homogenous with RCM clod (50.8%). Integrating dermoscopic pattern with contour, diameter, color, and anatomic location of nevi improved prediction of microscopic patterns beyond dermoscopy alone. The dermoscopic clinical regression model correctly classified lesions to RCM ring versus RCM clod in 90% and to RCM ring versus RCM composite patterns in 81%. Limitations The study was restricted to adults, back and lower extremities, and typical nevi. Conclusions Integrating dermoscopic patterns with clinical attributes may improve prediction of microscopic patterns of nevi.

Time reversibility is a fundamental hypothesis in time series. In this paper, Gini-based equivalents for time series concepts that enable to construct a Gini-based test for time reversibility under merely first-order moment assumptions are developed. The key idea is that the relationship between two variables using Gini (as measured by Gini autocorrelations and partial autocorrelations) can be measured in two directions, which are not necessarily equal. This implies a built-in capability to discriminate between looking at forward and backward directions in time series. The Gini creates two bi-directional Gini autocorrelations (and partial autocorrelations), looking forward and backward in time, which are not necessarily equal. The difference between them may assist in identifying models with underlying heavy-tailed and non-normal innovations. Gini-based test and Gini-based correlograms, which serve as visual tools to examine departures from the symmetry assumption, are constructed. Simulations are used to illustrate the suggested Gini-based framework and to validate the statistical test. An application to a real data set is presented.

This paper reports on the results of a drivers' survey regarding the effects of speed cameras for speed enforcement in Israel. The survey was part of a larger study that accompanied the introduction of digital speed cameras. Speed camera deployment started in 2011, and till the end of 2013 twenty-one cameras were deployed in interurban road sections. Yearly surveys were taken between 2010 and 2013 in 9 gas stations near speed camera installation sites in order to capture drivers' opinions about speed and enforcement. Overall, 1993 drivers were interviewed. In terms of admitted speed behavior, 38% of the drivers in 2010, 21% in 2011, 13% in 2012 and 11% in 2013 reported that their driving speed was above the perceived posted speed limit. The proportion of drivers indicating some speed camera influence on driving decreased over the years. In addition, the majority of drivers (61%) predicted positive impact of speed cameras on safety. This result did not change significantly over the years. The main stated explanation for speed limit violations was time pressure, while the main stated explanation for respecting the posted speed was enforcement, rather than safety concerns. Linear regression and sigmoidal models were applied to describe the linkage between the reported driving speed (dependent) and the perceived posted speed (independent). The sigmoidal model fitted the data better, especially at high levels of the perceived posted speeds. That is, although the perceived posted speed increased, at some point the actual driving speed levels off (asymptote) and did not increase. Moreover, we found that the upper asymptote of the sigmoidal model decreased over the years: from 113.22 (SE=18.84)km/h in 2010 to 88.92 (SE=1.55)km/h in 2013. A wide variance in perceived speed limits suggest that drivers may not know what the speed limits really are.

A meta-analysis was conducted to study the effect of a change in the speed limit on mean driving speed and safety on interurban roads. Information from 108 research results—documented in a database collected by Elvik et al. in 2004—was analyzed. The link between accident ratio (after–before) and the mean driving speed ratio was studied by estimating the power parameter in the Nilsson power model. A modified power model was applied to estimate the link between speed limit (rather than mean driving speed) ratio and accidents ratio. When the speed limit was raised, the mean driving speed ratio (mean driving speed after to mean driving speed before) was on average 1.04 (SD = 0.04), significantly higher than 1 but smaller than the speed limit ratio (average = 1.17; SD = 0.05). The power parameters in Nilsson’s power model and in its modified version were 3.73 (SD = 2.61) and 0.99 (SD = 0.46), respectively. When speed limit was reduced, the mean driving speed ratio was 0.93 (SD = 0.04), significantly smaller than 1 but larger than the speed limit ratio (average = 0.82; SD = 0.04). The corresponding power parameters were 2.81 (SD = 0.65) and 1.24 (SD = 0.44). This paper summarizes research on the link between changes in speed limit, driving speed, and safety. It was found that the average speed changed in the same direction as the speed limit, but the magnitude of the change was smaller. As a result, the power parameters in the second version of the power model were lower than those in the original Nilsson’s model, and so the average change in safety is nearly proportional to the change in speed limit.

BACKGROUND Vascular risk factors and lack of formal education may increase the risk of Alzheimer's disease (AD). OBJECTIVE To determine the contribution of vascular risk factors and education to the risk of mild cognitive impairment (MCI) and AD and to estimate the risk for conversion from MCI to AD. METHODS This door-to-door survey was performed by an Arab-speaking team in Wadi Ara villages in Israel. All consenting residents aged ≥ 65 years were interviewed for medical history and underwent neurological and cognitive examinations. Individuals were cognitively classified as normal (CN), MCI, AD, vascular dementia, or unclassifiable. MCI patients were re-examined at least one year later to determine conversion to AD. The contributions of age, gender, school years, and vascular risk factors to the probability of conversion were estimated using logistic regression models. RESULTS Of the 906 participants, 297 (33%) had MCI and 95 (10%) had AD. Older age (p = 0.0008), female gender (p = 0.023), low schooling (p < 0.0001), and hypertension (p = 0.0002) significantly accounted for risk of MCI versus CN, and diabetes was borderline (p = 0.051). The risk of AD versus CN was significantly associated with age (p < 0.0001), female gender (p < 0.0001), low schooling (p = 0.004) and hypertension (p = 0.049). Of the 231 subjects with MCI that were re-examined, 65 converted to AD. CONCLUSIONS In this population, age, female gender, lack of formal education, and hypertension are risk factors for both AD and MCI. Conversion risk from MCI to AD could be estimated as a function of age, time interval between examinations, and hypertension.

DNA repair is a major mechanism for minimizing mutations and reducing cancer risk. Here, we present the development of reproducible and specific enzymatic assays for methylpurine DNA glycosylase (MPG) repairing the oxidative lesions 1,N6-ethenoadenine (εA) and hypoxanthine (Hx) in peripheral blood mononuclear cells protein extracts. Association of these DNA repair activities with lung cancer was determined using conditional logistic regression with specimens from a population-based case-control study with 96 lung cancer cases and 96 matched control subjects. The mean MPG-εA in case patients was 15.8 units/μg protein (95% CI 15.3-16.3), significantly higher than in control subjects-15.1 (14.6-15.5), *P = 0.011. The adjusted odds ratio for lung cancer associated with a one SD increase in MPG-εA activity (2.48 units) was significantly bigger than 1 (OR = 1.6, 95% CI = 1.1-2.4; *P = 0.013). When activity of OGG1, a different DNA repair enzyme for oxidative damage, was included in the model, the estimated odds ratio/SD for a combined MPG-εA-OGG1 score was 2.6 (95% CI 1.6-4.2) *P = 0.0001, higher than the odds ratio for each single assay. The MPG enzyme activity assays described provide robust functional risk biomarkers, with increased MPG-εA activity being associated with increased lung cancer risk, similar to the behavior of MPG-Hx. This underscores the notion that imbalances in DNA repair, including high DNA repair, usually perceived as beneficial, can cause cancer risk. Such DNA repair risk biomarkers may be useful for risk assessment of lung cancer and perhaps other cancer types, and for early detection techniques such as low-dose CT.

a b s t r a c t This paper explores In-Vehicle Data Recorders (IVDRs) information about the count of undesirable driving events (such as hard braking, lane changing, and sharp turning) of 148 individuals. The information was logged over three years and included time stamp information about the occurrence of undesirable driving events in each trip (N = 573,238). The objective was to gain deeper understanding about the heterogeneity among drivers with respect to behavior change over time, the effect of trip duration and the distribution of events count. Our findings show that in some respects drivers are similar: for all drivers, the variance of the events count was larger than the mean, indicating that the negative binomial distribution is suitable to model the distribution of events count per trip. Most drivers (95%) had lower events rate during longer trips, suggesting that a ‘simple’ events rate index is problematic when comparing between those driving longer trips and drivers driving short trips. In addition, most drivers (87%) improved their driving behavior throughout the measurement period. However, there are important differences among drivers in terms of the frequency of behavior change and the trends in behavior over time. These findings demonstrate the need for personalized examination of individual drivers. Several tools for such personalized examination were developed and discussed in this study.

Only a minority of smokers develop lung cancer, possibly due to genetic predisposition, including DNA repair deficiencies. To examine whether inter-individual variations in DNA repair activity of N-methylpurine DNA glycosylase (MPG) are associated with lung cancer, we conducted a blinded, population-based, case–control study with 100 lung cancer case patients and 100 matched control subjects and analyzed the data with conditional logistic regression. All statistical tests were two-sided. MPG enzyme activity in peripheral blood mononuclear cells from case patients was higher than in control subjects, results opposite that of 8-oxoguanine DNA glycosylase (OGG1) DNA repair enzyme activity. For lung cancer associated with one standard deviation increase in MPG activity, the adjusted odds ratio was 1.8 (95% confidence interval [CI] = 1.2 to 2.6; P = .006). A combined MPG and OGG1 activities score was more strongly associated with lung cancer risk than either activity alone, with an odds ratio of 2.3 (95% CI = 1.4 to 3.6; P < .001). These results form a basis for a future panel of risk biomarkers for lung cancer risk assessment and prevention.

The prevalence of mild cognitive impairment (MCI) and Alzheimer's disease (AD) have not been well been studied in Arab populations. In a door-to-door study of all residents aged ≥ 65 years in Wadi-Ara, an Arab community in northern Israel, we estimated the prevalence of AD, MCI, and the risk of conversion to AD. Subjects were classified as cognitively normal, MCI, AD, or other based on neurological and cognitive examination (in Arabic). MCI subjects were re-examined (interval ≥ 1 year) to determine conversion to AD and contributions of age, gender, and education to the probability of conversion. Of the 944 participants (96.6% of those approached; 49.4% men), 92 (9.8%) had AD. An unusually high prevalence of MCI (n = 303, 32.1%) was observed. Since the majority of women (77.2%) had no schooling, we estimated the effect of gender on the risk of AD and MCI among subjects without schooling and of school years among men. Among subjects with no schooling (n = 452), age (p = 0.02) and female gender (p < 0.0001) were significant predictors of AD, whereas risk of MCI increased only with age (p = 0.0001). Among men (n = 318), age increased the risk (p < 0.0001), school years reduced the risk of AD (p = 0.039) and similarly for MCI [age (p = 0.0001); school years (p = 0.0007)]. Age (p = 0.013), but not gender or school years, was a significant predictor of conversion from MCI to AD (annual rate 5.7%). The prevalence of MCI and AD are unusually high in Wadi Ara, while the rate of conversion from MCI to AD is low. Yet unidentified genetic factors might underlie this observation.

Objective: In view of the fact that cancer patterns in patients with Parkinson disease (PD) differ from the general population, we aimed to verify whether patients with PD with LRRK2 mutations have an increased risk for particular cancer types. Methods: In this cross-sectional study, eligible consenting Jewish patients with PD were genotyped for the predominant LRRK2 G2019S mutation. Oncologic data were obtained by personal interview and reviewing patients9 files. Stepwise logistic regression was applied to model the probability of cancer occurrence in carriers vs noncarriers. Results: Overall, 79/490 (16.1%) genotyped patients carried the G2019S mutation. Seventy-seven (16%) were diagnosed with cancer; of those, 67 (14%) with a non-skin cancer. Eighteen (23%) carriers vs 49 (12%) noncarriers had a non-skin cancer ( p = 0.01, odds ratio [OR] = 2.18, 95% confidence interval [CI] 1.19–3.99). A significant ethnicity effect was noted ( p = 0.045, OR = 1.84, 95% CI 1.02–3.34). Among Ashkenazi patients, age and LRRK2 emerged as significant using stepwise logistic regression including age, gender, and LRRK2 status as explanatory variables. The OR for LRRK2 mutation carriers adjusted for age was 3.38 (95% CI 1.64–6.97, p = 0.0009). Conclusions: Ashkenazi Jewish patients with PD who harbor the G2019S LRRK2 mutation are more likely to have a concomitant non-skin cancer than noncarriers.

There is a perceived tradeoff between the ease of measuring alcohol in the body and the accuracy of the result. Direct tests of blood alcohol concentrations are considered the most accurate, desktop stationary breath testers based on electro-chemical infra-red technology are slightly less accurate, but accepted for evidentiary purposes in most jurisdictions, and quick portable breath testers based on fuel-cell technology are the easiest to administer but not acceptable in many courts. This study compared the accuracy of an evidentiary portable breath tester and an evidentiary desktop breath tester relative to blood alcohol concentrations. Inverse regressions were used to obtain confidence limits for the alcohol levels as read by the breath testers that would provide tradeoffs of false positives and false negatives for three levels of confidence: 95%, 96%, and 98%; corresponding to false positive values of 2.5%, 2%, and 1%, respectively. A decision tree model is offered for the optimal use of the three measures, so that portable breath testers can be sufficient for high level BrAC, stationary breath testers can be sufficient for medium level BrAC, and blood tests are recommended for still lower BrACs. The model provides quantitative BrAC threshold levels for the two most common BAC levels used to imply DWI: 50 mg/dl and 80 mg/dl.

Mild Cognitive Impariment (MCI) and healthy aging have been shown to be associated with Mild Parkinsonian Signs (MPS). We performed a door-to-door observational and follow-up study amongst consenting residents of Wadi Ara Arab villages in northern Israel aged ≥ 65 years (n = 687) to examine whether MPS represent a risk factor for MCI and/or conversion from MCI to Alzheimer’s disease (AD). In Phase 1, 223 cognitively normal (CN) and 173 MCI subjects were assessed by interview for medical history, neurological examination, motor part of the Unified Parkinson Disease Rating Scale (mUPDRS) (divided into item-clusters: axial, limb bradykinesia, tremor and rigidity) and cognitive tests. MCI subjects (n = 111) were reevaluated in Phase 2 for conversion to AD at least one year after initial assessment. MCI subjects had a higher frequency of axial dysfunction (8.7% vs. 1.3%) and limb bradykinesia (10.4% vs. 1.3%) than CN subjects (p < 0.001, both). Stepwise logistic regression analysis estimating the probability of MCI vs. CN revealed higher mUPDRS (OR = 1.19, 95% CI, 1.05 to 1.35, p = 0.006) and higher limb bradykinesia scores (OR = 1.75, 95% CI, 1.2 to 2.56, p = 0.003) and not age as explanatory variables. Presence of MPS did not predict conversion to AD after adjustment for age and time-interval. These results suggest that axial and bradykinetic parkinsonian signs represent risk factors for MCI but MPS may not predict conversion from MCI to AD.

The cause of the majority of road crashes can be attributed to drivers' behavior. Recent in-vehicle monitoring technologies enable continuous and high resolution measurements of drivers' behaviors. We analyzed the information received from a novel in-vehicle technology which identifies the occurrences of undesirable driving events such as extreme braking and accelerating, sharp cornering and sudden lane changing. We undertook an exploratory analysis to provide better understanding of events frequency (EF) statistical properties. Our findings show higher EF in trip edges (trip beginning and trip end) than in the middle of the trip, higher EF for males than for females and higher EF at nighttime than at daytime. Use of the in-vehicle technology's continuous and high resolution measurements enabled interesting advanced statistical analyses. Future research can use our findings to build similar statistical models to predict the occurrence of undesirable driving events by other independent variables.

Background: We aimed to determine whether vascular risk factors are associated with Alzheimer’s disease (AD) and mild cognitive impairment (MCI) in an elderly Arab population. Methods: An Arabic-speaking team performed a door-to-door survey of consecutive residents aged ≧65 years. We estimated the odds of AD or MCI versus normal controls as a function of age, gender, education and presence of vascular factors by multinomial logistic regression with interactions. Results: Out of 767 subjects (54% men), 444 were cognitively normal, 234 had MCI and 89 had AD. AD was significantly associated with hypertension (p = 0.01; OR = 2.08; 95% CI: 1.18–3.65), age (p < 0.0001; OR = 1.19; 95% CI: 1.14–1.24), female gender (p = 0.0016; OR = 3.06; 95% CI: 1.53–6.15) and education (p = 0.0002; OR = 0.75; 95% CI: 0.65–0.88). MCI was significantly associated with hypertension (p = 0.0042; OR = 1.69; 95% CI: 1.25–2.44), age (p < 0.0001; OR = 1.06; 95% CI: 1.03–1.09) and education (p < 0.0001; OR = 0.76; 95% CI: 0.71–0.83), but not with gender. Conclusions: Hypertension, older age and low education significantly increase the probability of AD and MCI. The effect of hypertension on the odds of AD versus controls is over and above the effects of age, gender and education. For MCI versus controls there is no gender effect, and the effect of hypertension is over and above the effects of age and education.

The huge technological improvement in data processing and the globalization have increased the demand for and the supply of indices that quantify the consequences of a policy. However, there are certain cases in which quantification may be misleading in the sense that it gives the impression of an accurate measurement while in reality it is not. Needless to say that a policy that is based on the wrong impression is doomed to fail. This article presents such an example in the area of education. Ability is a latent variable. As such, it has to be measured indirectly, using questions of varying levels of difficulties. The larger the number of questions a student answers correctly, the higher the grade and the higher the estimated ability. Groups (i.e., classes, schools, ethnic groups, etc.) are ranked in a similar way, based on the average grades of their students. This article argues that the inability to directly measure ability may cause the ranking of groups to be questionable in the sense that one can c...

How does practice make perfect in the acquisition of morphological skill in adults? Participants underwent intensive, multi-session training on an artificial morphological rule (AMR) requiring phonological transformations of verbs according to a semantic distinction. All participants learned to apply the AMR to repeated items, with a power law like improvement in speed and accuracy (group average), both within-sessions and between-sessions (consolidation phase) gains, and robust retention, as in non-linguistic skill learning. Generalization to new items evolved separately for different aspects of the AMR. Phonological aspects were generalized by all participants, independently of explicit (declarative) knowledge, and well fitted by a power function. However, the generalization of the semantic aspect required the explicit discovery of its requisite role, and was not universally attained; when attained, explicit knowledge of the semantic aspect of the AMR coincided with an abrupt increase in accuracy and initiated a phase of fluency gains (proceduralization). Our results suggest that both procedural and declarative memory contribute differentially to the learning of distinct aspects of a morphological rule, at different stages along the mastering of skilled linguistic performance.

This paper presents a methodology for detailed evaluation of autonomous automated guided vehicles systems (AGVS) used for material handling. The methodology includes: stand-alone sub-module evaluation, including comprehensive simulations and statistical analysis of the system's sub-modules, along with hardware validation; quantitative system evaluation for integrated system performance investigation; and structured qualitative analyses for identifying strengths and weaknesses not readily apparent. The defined performance measures include aspects from both multi-robot and AGV fields. The developed methodology provides a systematic way to model, experiment with, analyze, and compare different AGVS control methods. To demonstrate the methodology, it was applied to evaluate a recently developed decentralized AGVS control method.

A nonparametric multiple regression method, based on the extended Gini that depends on one parameter, v, is investigated. The parameter v enables production of infinite alternative linear approximations to the regression curve which differ in the weighting schemes applied to the slopes of the curve. The method allows the investigator to stress different sections of one independent variable while keeping the treatment of the other independent variables intact. As an application we investigate nonresponse patterns in a survey of household expenditures to learn about the relationship between nonresponse and income. The empirical results show that the higher the income, the higher the response rate, and the larger the household, the higher the response rate. The Arab population tends to respond more than the Jewish one, whereas the ultrareligious group tends to respond less than the rest of the population. The implications on the bias in the estimates are discussed.

We consider statistical tests concerning various relationships between two absolute concentration curves (ACCs). In particular, we consider tests for determining if the two ACCs coincide, if one is above another in a specified order, or if they do not intersect without specifying which one is above/below the other one. These problems are of interest in the context of marginal conditional stochastic dominance (MCSD). Constructing statistical tests for the MCSD relies on ideas and also on their modifications developed by Linton, Maasoumi, and Whang (2005, Review of Economic Studies 72, 735–765) in the context of stochastic dominance for distribution functions. Our theoretical considerations are supplemented with a simulation study.

The connection between the area under the ROC curve (AUC), which is frequently used in the diagnosis and classification literature, and the Gini terminology, which is mainly used in the economic literature, is clarified. It is shown that AUC is related to the covariance between Yi, the number of 1's until the ith 0, and F(Ti), the empirical rank of the ith 0, ordered by the predictive probability.

The primary objective is to compare between statistical and SPC methods for monitoring process variance via an extensive simulation. The statistical methods are Cumulative Sum (CUSUM) and Likelihood Ratio Test (LRT), and the SPC methods are Moving Range (MR) and Exponentially Weighted Mean Squared deviation (EWMS). In addition, we examined a combination of both statistical and SPC methods, denoted by “Schechtman on S2”. Random series were generated from three distributions: normal, Poisson and lognormal. The parameters of interest were: series length, size of the change, and location. The criteria for comparison were: rates of false alarms, powers and Mean Absolute Deviation (MAD). The main outcomes and results of our simulations show that when the distribution is normal, all procedures are slightly conservative. However, when the distribution is not normal, the rates of false alarm go up to around 0.8 for all but Schechtman on S2. Assuming normal distribution, EWMS and CUSUM are powerful for detecting ch...

We consider the two-sample t -test where error variances are unknown but with known relationships between them. This situation arises, for example, when two measuring instruments average different number of replicates to report the response. In particular we compare our procedure with the usual Satterthwaite approximation in the two sample t -test with variances unequal. Our procedure uses the knowledge of a known ratio of variances while the Satterthwaite approximation assumes only that the two variances are unequal. Simulations show that our procedure has both better size and better power than the Satterthwaite approximation. Finally, we consider an extension of our results to the General Linear Model.

Background Vitreous hemorrhage (VH) is a major cause of severe vision loss in diabetic patients. The aim of this study was to assess the incidence and risk factors for new VH in diabetics previously treated with panretinal photocoagulation (PRP) for proliferative retinopathy (PDR) in community base center. Methods Records of 192 diabetics (35 type 1, 157 type 2), undergoing PRP for diabetic retinopathy were retrospectively reviewed. Eyes presenting initially with high-risk PDR received PRP without delay, and eyes presenting initially with severe non proliferative retinopathy (NPDR) or early PDR had undergone central retinal photocoagulation (CRP), and then, when high risk PDR developed, received PRP treatment. Results VH had developed in 39% of the eyes despite PRP. Risk factors for VH in type 1 diabetes were: early onset and long duration of disease (23.8 versus 39.0 years of age, P = 0.007, and 25.8 versus 16.0 years, P = 0.002, respectively). In type 2, VH occurred with less follow-up and angiographic examinations (7.4% versus 3.8%, P = 0.027, and 33% versus 47%, P = 0.07, respectively). CRP decreased the risk for VH from 43 to 15%, P = 0.013. Conclusions In type 2 diabetes, regular ophthalmic follow-up and intensive PRP may reduce the risk for VH in eyes previously treated by PRP. In type 1, early onset disease and long duration are associated with higher incidence of VH.

Quantiles are frequently used as descriptive measures. When data contains measurement errors, using the contaminated data to estimate the quantiles results in biased estimates. In this paper, we suggest two methods for reducing the effect of measurement errors on the quantile estimates and compare them, via an extensive simulation study, to the estimates obtained by the naive method, that is: by the estimates obtained from the observed (contaminated) data. The method we recommend is based on a method in a paper by Cook and Stefanski. However, we suggest using a combination of bootstrap and jackknifing to replace their extrapolation step.

While the role of reduced DNA repair in susceptibility to hereditary cancers is well established, its role in sporadic cancer is less understood. One of the reasons is the lack of specific DNA repair assays that are suitable for epidemiology studies. Here we describe the development of the OGG test, an epidemiology-grade enzymatic assay for the activity of the base excision repair enzyme 8-oxoguanine DNA glycosylase, in protein extracts prepared from human blood cells. The assay is robust and reproducible, with a coefficient of variation of 10%. Using the OGG test we determined OGG activity in 120 healthy individuals. Our results show an inter-individual variation of 2.8-fold in OGG activity, from 3.6 up to 10.1 units/μg protein, with a mean value of 7.2 units/μg protein. There was no significant difference in OGG activity between males and females, or between smokers and non-smokers. Interestingly, there was a gender-specific effect of age: OGG activity was slightly but significantly lower in males older than the age of 55 years compared to younger males, but not in females at the same age groups. Analysis of OGG1 mRNA by quantitative real-time RT-PCR showed a group trend of an increase in OGG enzymatic activity with increasing mRNA expression, but the correlation between activity and mRNA in individuals was poor, indicating the importance of factors other than mRNA expression. The OGG test described is expected to be useful in studying the role of 8-oxoguanine repair in cancer, as recently demonstrated for non-small cell lung cancer [T. Paz-Elizur, M. Krupsky, S. Blumenstein, D. Elinger, E. Schechtman, Z. Livneh, J. Natl. Cancer Inst. 95 (2003) 1312–1319]. In addition, it may serve as a paradigm for the development of additional functional DNA repair tests, which are needed in order to gain further insight into the role of DNA repair in cancer risk and pathology.

An increasing number of studies indicate that reduced DNA-repair capacity is associated with increased cancer risk. Using a functional assay for the removal of the oxidative DNA lesion 8-oxoguanine by the DNA-repair enzyme 8-oxoguanine DNA glycosylase 1 (OGG1), we have previously shown that reduced OGG activity is a risk factor in lung cancer. Here, we report that OGG activity in peripheral blood mononuclear cells from 37 cases with squamous cell carcinoma of the head and neck (SCCHN) was significantly lower than in 93 control subjects, frequency matched for age and gender. Retesting of OGG activity 3 to 4 years after diagnosis and successful treatment of 18 individuals who recovered from the disease showed that OGG activity values were similar to those determined at diagnosis, suggesting that reduced OGG activity in case patients was not caused by the disease. Logistic regression analysis indicated that the adjusted odds ratio (OR) associated with a unit decrease in OGG activity was statistically significantly increased [OR, 2.3; 95% confidence interval (95% CI), 1.5–3.4]. Individuals in the lowest tertile of OGG activity exhibited an increased risk of SCCHN with an OR of 7.0 (95% CI, 2.0–24.5). The combination of smoking and low OGG was associated with a highly increased estimated relative risk for SCCHN. These results suggest that low OGG is associated with the risk of SCCHN, and if confirmed by additional epidemiologic studies, screening of smokers for low OGG activity might be used as a strategy for the prevention of lung cancer and SCCHN. (Cancer Res 2006; 66(24): 11683-9)

A wildly discussed shortcoming of panel surveys is a potential bias arising from selective attrition. Based on data of the German Socio-Economic Panel Study (SOEP), the authors analyze potential artifacts (level, structure, income inequality) by comparing results for two independently drawn panel subsamples started in 1984 and 2000. They apply ANOGI (analysis of Gini) techniques, the equivalent of ANOVA performed with the Gini coefficient. They rearrange, reinterpret, and use the decomposition in the comparison of subpopulations from which the different samples were drawn. Taking into account indicators for income, significant differences between these two samples with respect to income inequality are found in the first year, which start to fade away in Wave 2 and disappear in Wave 3. The authors find credible indication for these differences to be driven by changes in response behavior of short-term panel members rather than by attrition among members of the longer running sample. Dieser Beitrag ist mit Zustimmung des Rechteinhabers aufgrund einer (DFG-geförderten) Allianz- bzw. Nationallizenz frei zugänglich. - This publication is with permission of the rights owner freely accessible due to an Alliance licence and a national licence (funded by the DFG, German Research Foundation) respectively.

In this study, we propose formal models and algorithms to detect drug impairment and identify the impairing drug type, on the basis of data obtained by a Drug Evaluation and Classification (DEC) investigation. The DEC program relies on measurements of vital signs and observable signs and symptoms. A formal model, based on data collected by police officers trained to detect and identify drug impairments, yielded sensitivity levels greater than 60% and specificity levels greater than 90% for impairments caused by cannabis, alprazolam, and amphetamine. For codeine, with a specificity of nearly 90% the sensitivity was only 20%. Using logistic regression, the formal model was much more accurate than the trained officers in identifying impairments from cannabis, alprazolam, and amphetamine. Both the formal model and the officers were quite poor in identifying codeine impairment. In conclusion, the joint application of the DECP procedures with the formal model is useful for drug detection and identification.

A double blind study was performed to evaluate the ability of trained police officers to detect drug impairments and to identify the type of drug responsible for the impairment, on the basis of observed symptoms and psychophysical measurements of performance alone. The officers were not allowed to interview the subjects, and their conclusions were based solely on observable signs and symptoms on systematically measured vital signs, and on standardized sobriety tests of motor coordination. Results showed that with this partial information, the officers are able to detect drug impairment at better-than-chance levels with a sensitivity (correct detection of impairments) of 72%, but with a specificity of 43% (false alarm rate of 57%). Furthermore, the association between drug ingestion and identification of the specific impairing drug category was not very high, with sensitivities ranging from a low of 10% for amphetamine to a high of 49% for cannabis. Based on both sensitivity and specificity, drug identification was best for alprazolam impairment, noticeably poorer for cannabis and codeine impairment, and no better than chance for amphetamine impairment. Performance could have been improved if the officers were to list the two most probable impairing drugs (rather than limit their decision to only one), and if they were more consistent in their interpretation of observable signs and symptoms.

This paper examines the effect of the use of Center High Mounted Stop Lamp (CHMSL) on rear-end accidents, as reflected in Israeli police records from calendar years 1991-2002. The basic analysis, similar to that used in previous CHMSL studies, compares the involvement in accidents of passenger cars of model years 1994-1996 that are equipped with CHMSL with passenger cars of model years 1991-1993 that are not equipped with CHMSL. The number of involvements as the struck vehicle in a rear-end accident was used as the relevant measurement and the number of involvements as the striking vehicle in a rear-end accident was used as the reference measurement. The results yielded an odds ratio of 0.93. The explanation that the CHMSL is responsible for the 7% decrease is intuitively appealing and is consistent with previous findings. However, the strength of this evidence is marginal (p = 0.07). Additional analyses evaluated the model year effect in greater detail, in order to determine whether there exists a change point between 1993 and 1994 as would be expected from a CHMSL effect, or whether the effect is spurious. Detailed analyses were performed on the ratio of struck to striking involvements as well as the rates of involvement of both types. These analyses showed that (1) the chosen reference measurement is an appropriate one, but (2) the 0.93 odds ratio is quite possibly due to other reasons unrelated to the CHMSL, thus further limiting the confidence in CHMSL effectiveness.

In this paper, a new simple method for jackknifing two-sample statistics is proposed. The method is based on a two-step procedure. In the first step, the point estimator is calculated by leaving one X (or Y ) out at a time. At the second step, the point estimator obtained in the first step is further jackknifed, leaving one Y (or X ) out at a time, resulting in a simple formula for the proposed point estimator. It is shown that by using the two-step procedure, the bias of the point estimator is reduced in terms of asymptotic order, from O( n −1 ) up to O( n −2 ), under certain regularity conditions. This conclusion is also confirmed empirically in terms of finite sample numerical examples via a small-scale simulation study. We also discuss the idea of asymptotic bias to obtain parallel results without imposing some conditions that may be difficult to check or too restrictive in practice.