Please activate JavaScript!
Please install Adobe Flash Player, click here for download


ISCB 2014 Vienna, Austria • Abstracts - Poster Presentations 139Wednesday, 27th August 2014 • 15:30-16:00 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust P4.5.141 “Numberstories”: Teaching statistics to kids - and clinicians J Røislien1,2 , KF Frøslie1,3 1 Department of Biostatistics, University of Oslo, Oslo, Norway, 2 Department of Health Studies, University of Stavanger, Stavanger, Norway, 3 Norwegian Resource Centre Women’s Health, Oslo Uni. Hospital, Oslo, Norway   The use of statistics has revolutionized medicine, and statistical analysis is at the core of evidence based medicine and the communication of medi- cal research to the general public. However, most non-statisticians strug- gle with the mathematical language of statistics. Indeed, people tend to use statistics like they use their cars: They’re happy if the user interface is fine, if it works like it is supposed to, and takes them where they want to go. People generally care just as little about what goes on under the hood of their cars, as do clinicians about formulas and algorithms. Still, they get their driver’s license. Still, they drive cars. “Tall forteller” (Eng: “Numberstories”) takes a similar approach to teaching statistics. Awash with pop-art photos and anecdotal introductions, focus is on statistics as a vehicle for analyzing quantitative information, rather than mathematical formulae. Through nineteen simple research projects, which the reader is encouraged to perform herself, she is - unknowingly - guided through a basic course in descriptive statistics and bivariate analy- sis. The text is simple and can be understood by anyone having their first introduction to statistics; from teenagers to medical doctors. Common statistical terms are explained in plain language. Formulas are replaced by a free app performing all calculations necessary to answer the nineteen research questions, presenting graphs and statisti- cal tests at the push of a button, given that the reader knows the type and number of variables. “Tall forteller” teaches how to drive the statistics car. P4.5.158 A model of population dynamics based on fuzzy cellular automata theory A Szymański1 , A Rossa1 1 Institute of Stat. and Demography, University of Lodz, Lodz, Poland   In the paper a hybrid model based on cellular automata and fuzzy logic is proposed to simulate the dynamics of a population, where the changes in fertility and mortality rates are induced by a stochastic variation of the environment. The concept of the Cellular Automata (CA) is based on the original von Neumann’s idea (1966), modified by Betel et al. (2011) in a form of the so- called fuzzy cellular automata (FCA). CA is a formal model composed of a rectangular region of MxN cells in which the evolution of each cell de- pends on its present state and a state of its neighboring cells. All the cells pass through the following generation at the same time, according to the transition function being common for all the cells. In the proposed model the environment variables are treated as fuzzy sets. Their combination is treated as a result of factors determining population growth. The operations on fuzzy sets lead to fertility and mortality rates. The obtained results have been verified by means of the mortality data for Poland. This research was supported for both authors by a grant from the National Science Center under contract DEC-2011/01/B/HS4/02882, for which the authors are indebted.   P4.5.159 Studying latent hydration level through multivariate longitudinal biological indicators: application of a latent process mixed model J Tanguy1 , C Proust-Lima2,3 1 Danone R.D., Palaiseau Cedex, France, 2 INSERM U897, Epidemiology and Biostatistics Research Center, Bordeaux, France, 3 University of Bordeaux, ISPED, Bordeaux, France   Clinicians are often faced with the problem of not being able to measure directly what they want to study. In this case, multivariate longitudinal variables measuring the latent construct of interest are collected, and the aim is to establish causal-effect relationship between explanatory covari- ates and this latent concept. In addition, understanding how outcomes are related to each other and with respect to their underlying concept can also be relevant. To address these challenges, an appealing solution is to use latent process mixed models assuming that each longitudinal indicator (whatever its na- ture) is a noisy measure of the same underlying function. In short, measurement models link the observed outcomes (quantita- tive, ordinal,...) to the latent concept, and a structural model describes the latent concept according to explanatory covariates. This model borrows strength across outcomes while taking into account the correlation within each outcome over time and between outcomes. This work aims at illustrating the possibilities offered by this approach through a study case on the issue of measuring hydration in real-life. Hydration can be measured by multiple biological measures from urines or plasma and by total fluid and water intake. However, it is not clear which markers are good proxies of the absolute hydration level. Yet, identifying proxies of hydration becomes essential to quantify hydration in popula- tions and evaluate optimal levels of hydration. Goodness-of-fit, interpretation, limits and advantages of the method, but also perspectives with respect to regulatory requirements and specific challenges in clinical studies in nutrition are discussed.   P4.5.165 Item analysis and evaluation of the examinations in faculty of medicine L Tomak1 , Y Bek1 1 Ondokuz Mayis, Samsun, Turkey Item analysis is an effective method in the evaluation of examinations.This study aims the comparative evaluation of the classical and latent methods used in Item Analysis, and the efficacy of these methods in the evaluation of the examinations of Faculty of Medicine. The student exams in the Faculty of Medicine were evaluated using dif- ferent methods. The methods to be used can be divided in two main classes; classical methods and latent methods. Among the classical meth- ods, Cronbach´s alpha, split half methods, item discrimination and item difficulty were investigated. On the other hand, various models of item response theory and their statistics were compared in the group of latent methods. Different statistics and measurements obtained by the application of classical and latent methods were evaluated, and we tried to determine “which method is more reliable under which evaluation criteria”. The most ideal evaluation criteria, evaluation method and the software were deter- mined by evaluating the implementation results of the computer soft- wares developed for this purpose.  

Pages Overview