Kazim Ozdamar

Comparing classification techniques for predicting essential hypertension

Hypertension is a leading cause of heart disease and stroke. In this study, performance of classification techniques is compared in order to predict the risk of essential hypertension disease. A retrospective analysis was performed in 694 subjects (452 patients and 242 controls). We compared performances of three decision trees, four statistical algorithms, and two neural networks. Predictor variables were age, sex, family history of hypertension, smoking habits, lipoprotein (a), triglyceride, uric acid, total cholesterol, and body mass index (BMI). Classification techniques were grouped using hierarchical cluster analysis (HCA). The data points appeared to cluster in three groups. The first cluster included MLP and RBF. Furthermore CART which was more similar than other techniques linked this cluster. The second cluster included FDA/MARS (degree=1), LR and QUEST, but FDA/MARS (degree=1) and LR was more similar than QUEST. The third cluster included FDA/MARS (degree=2), CHAID and FDA, but FDA/MARS (degree=2) and CHAID was more similar than FDA. MLP and RBF which are one each of neural networks procedures, performed better than other techniques in predicting hypertension. QUEST had a lesser performance than other techniques.

Comparison of Semiparametric, Parametric, and Nonparametric ROC Analysis for Continuous Diagnostic Tests Using a Simulation Study and Acute Coronary Syndrome Data

We aimed to compare the performance of three different individual ROC methods (one from each of the broad categories of parametric, nonparametric and semiparametric analysis) for assessing continuous diagnostic tests: the binormal method as a parametric method, an empirical approach as a nonparametric method, and a semiparametric method using generalized linear models (GLM). We performed a simulation study with various sample sizes under normal, skewed, and monotone distributions. In the simulations, we used estimates of the ROC curve parameters a and b, estimates of the area under the curve (AUC), the standard errors and root mean square errors (RMSEs) of these estimates, and the 95% AUC confidence intervals for comparison. The three methodologies were also applied to an acute coronary syndrome dataset in which serum myoglobin levels were used as a biomarker for detecting acute coronary syndrome. The simulation and application studies suggest that the semiparametric ROC analysis using GLM is a reliable method when the distributions of the diagnostic test results are skewed and that it provides a smooth ROC curve for obtaining a unique cutoff value. A sample size of 50 is sufficient for applying the semiparametric ROC method.

Comparison of Bayesian survival analysis and Cox regression analysis in simulated and breast cancer data sets

We aimed to compare the performance of Cox regression analysis (CRA) and Bayesian survival analysis (BSA) by using simulations and breast cancer data. Simulation study was carried out with two diff...

Sex-related differences in methods of and reasons for suicide in Turkey between 1990 and 2010

Objective To analyse suicides occurring in Turkey between 1990 and 2010 according to sex, age, reason and method. Methods Data concerning suicides occurring in Turkey between 1990 and 2010 were obtained from annual records published by the Turkish Statistical Institute. Data were analysed according to sex, age, reason for suicide and suicide method. Results The crude suicide rate was 2.42 per 100 000 population in 1990 and 4.02 per 100 000 population in 2010. The highest number of suicides was seen in those aged 15–24 years. In this age group, the number of suicides in females was significantly higher than in males. The most common suicide method in Turkey was hanging. Men used firearms more frequently than women. All the reasons for suicide occurred in a higher number of males than females; however, the difference between males and females was more significant for economic problems, relationship problems and educational failure. The leading reason for suicide in females was relationship problems. Conclusion Detecting population subgroups with a high suicide risk and obtaining regional demographic data are of great importance for future studies on suicide prevention.

Bayesian Analysis of Parametric Survival Models: A Computer Simulation Study based Informative Priors

AbstractIn the article, the performance of Bayesian parametric survival models (Weibull, exponential, log-normal and log-logistic) by using Monte Carlo simulation was empirically compared while varying the informative priors and the sample sizes. We simulated the generated data by running for each of Weibull, exponential, log-normal and log-logistic survival models under varying informative priors and sample sizes using our simulation algorithm. For each situation, 1000 simulations were performed. Models with proper informative prior showed a good performance with too little bias. It was found out that bias of models increased while priors were becoming distant from reliability in all sample sizes. According to results obtained from simulation study, researchers should avoid assessment of data by using only one parametric survival model in future studies. We suggest that data should be better explored and processed by high performance modelling methods. Especially, quality of prior information to update k...Abstract In the article, the performance of Bayesian parametric survival models (Weibull, exponential, log-normal and log-logistic) by using Monte Carlo simulation was empirically compared while varying the informative priors and the sample sizes. We simulated the generated data by running for each of Weibull, exponential, log-normal and log-logistic survival models under varying informative priors and sample sizes using our simulation algorithm. For each situation, 1000 simulations were performed. Models with proper informative prior showed a good performance with too little bias. It was found out that bias of models increased while priors were becoming distant from reliability in all sample sizes. According to results obtained from simulation study, researchers should avoid assessment of data by using only one parametric survival model in future studies. We suggest that data should be better explored and processed by high performance modelling methods. Especially, quality of prior information to update knowledge about the parameters was a very important.

Violence in health institutions: a survey of health care workers in west Turkey.

To th e Editor: Violent, aggressive or abusive behaviour towards health care workers is an ever-increasing concern in today’s health care workplace.1 Some studies in connection with violence conducted in Turkey show that the frequency of violence has increased recently.23 Health care workers are particularly vulnerable due to the nature of their jobs, which include dealing with clients who may be emotionally disturbed.4 O ne of the most difficult situations that health care workers face is the threat o f physical harm at the hands o f a patient, a patients’ relatives, a colleague or superior.5 Job conditions such as working alone or in small numbers, working late at night, working in high-crime and unsecured areas, work load, job-related stress, role conflict, lack of supervisor support, low visibility, and poor lighting,6 8 which have been associated with workplace violence, may signal opportunity for the perpetrator. We conducted a study to learn more about the prevalence rates and risk factors for violence in various workplace conditions, providing a specific focus on the extent and nature of violence against health care workers. O ur study was carried out in the cities o f Eskisehir, Ankara, Kutahya and Bilecik, in the western part o f Turkey. The study population was comprised of staff at 34 health institutions. All 1368 health care workers working in the four study areas were recruited randomly during routine visits to local health institutions, and of these, 1209 workers agreed to participate. We evaluated the prevalence o f violence towards health care workers in a oneyear period, regardless o f the types of violence at the time o f the interview. This study was approved by Osmangazi University Management and the Turkish Medical Association. We used a Form of Violence Experience (FVE), a self-report questionnaire developed from the Violent Incident Form (VIF).9 Health care workers were asked to recall physical, sexual and emotional experiences o f violence directed at them at work over the previous year. The data were analyzed by a binary logistic regression model when appropriate to examine group disparities. The overall response rate was 88.4% (1209/1368). Ages of respondents ranged 16 to 73 years (mean, 32 years). Female health care workers constituted 38.9% of the total. Nearly 50% [n=598 (49.5%)] o f the respondents had experienced one or more types o f violence during the past year at work: 14.5% once, 33.8% between twice and five times, and 51.7% six or more times. The most frequent form of violence was emotional/verbal (72.4%) compared to physical (11.7%) and physicalemotional/verbal (15.9%). O f 598 violent events, 47.7% were reported in men, and 50.7% in women, which did not show a significant difference with regards to women. Compared with Eskisehir, which has the best socio-economic development, the risk of being exposed to violence in Bilecik, Kutahya and Ankara was reduced (odds ratio=0.46, 0.41, 0.75, respectively). Compared with specialist physicians, ancillary health personnel, dentists, pharmacists, security guards and academic staff were at a higher risk o f experiencing violence while the potential for exposure to violence was lower for assistant doctors, nurses and general practitioners. However, attendants experienced nearly the same risk of violence as that o f specialist doctors. Only emergency service was associated with significantly less exposure to violence compared with other services (7V0.05). Length of time working in the health care sector proved a risk factor in being exposed to violence (7V0.05). M ost aggressors were patient relatives/visitors (57.5%). The most common aggressors were men (76.9%), between 31 to 50 years o f age (55.3%), and those with a regular income (19.2%), not counting those for whom income was unknown (63.6%). Health care workers in this region of Turkey work in an environment in which they are constantly exposed to situations with aggressive/violent individuals. O ur findings are consistent w ith previous studies suggesting a 25% to 88% incidence of any form of violence in the previous year.13,10,11 The results presented in these studies show that workplace

The effect of different data structures, sample sizes on model fit measures

ABSTRACT In this study, our aim was to investigate the changes of different data structures and different sample sizes on the structural equation modeling and the influence of these factors on the model fit measures. Examining the created structural equation modeling under different data structures and sample sizes, the evaluation of model fit measures were performed with a simulation study. As a result of the simulation study, optimization and negative variance estimation problems have been encountered depending on the sample size and changing correlations. It was observed that these problems disappeared either by increasing the sample size or the correlations between the variables in factor. For upcoming studies, the choice of RMSEA and IFI model fit measures can be suggested in all sample sizes and the correlation values for data sets are ensured the multivariate normal distribution assumption.

Factors influencing students’ success: A generalized estimating equations study

The goals of this study were to determine the factors influencing children’s success according to Generalized Estimating Equations (GEE) through the Family Report Form (FRF; constituted by the authors), Teacher’s Report Form (TRF; Achenbach, 1991b) and Child Behavior Check List (CBCL; Achenbach & Edelbrock, 1983), and to form a new and actual reference success form for Turkish children. The study sample included 4130 primary school children in Eskisehir, Turkey. School success was measured as the total scores that students achieved in the following classes: cultural, social, science, foreign language, computer, picture, music and job-occupation. The TRF, CBCL and FRF forms consist of the sum of subtests, which includes 9 factor scales, namely anxiety/depression, somatic complaints, social withdrawal, delinquent behavior, social problems, thought problems, attention deficits, aggressive behavior, and other. As the statistical analyses, GEE method, stepwise regression analysis and missing value analyses were used. The items that had the greatest negative effect on the students’ success were “has poor school work” and “can’t concentrate or can’t pay attention for long”. On the other hand, the item that had the greatest positive effect on students’ success was “feels he/she has to be perfect”. It may be recommended that students, their families, and teachers should be informed about factors influencing school success.