Events in 2007

Superficially observing opinion polls, one could easily dismiss the polling results from the presidential election period as a failure of opinion surveys. The lecture will try to answer the question whether this was actually the case. We will see that in Slovenia, as well as in the neighbouring Croatia, very interesting things took place.

1. All the polls for the first round of presidential election one week before the election forecasted about 39% for Peterle, but the week later he only got 29% of the vote. Was this an error of the surveys or was the difference caused by the focused campaigns and "low blows" during the final week?
2. Never before have the 2nd and 3rd candidate (Gaspari and Türk) oscillated so much as this year, and never before have the polling days of the week and marginal events, news, articles etc. influenced the preferences so much.
3. In Croatia, the HDZ and SDP parties won wogether 85% of the vote, while the forecasts were 70% maximum. Never before have small parties lost so much with respect to the polls. Interestingly, the survey results played a part in that! The effect has a curious name - guillotine!
4. Hardly noticeable, but still: exit polls as a final survey of voting outcomes underestimated the HDZ for almost 3%. It is interesting to note that underestimation of the "right-wing" in exit polls is a general phenomenon, which has also been observed during the last presidential election in the USA and is also present to a lesser extent in Slovenia, but it has never before reached the Croatian proportion.
5. Last but not least, the polling/survey industry has never before been so explicitely under fire. Janša and Sanader alike openly attacked specific companies (Ninamedia in Slovenia, Puls in Croatia). The consequence is clear – at least in a part of the population, the politicians discredited the research. A potential problem is that such discrediting affects their voters more than others, so the response rate could thus become lower among those voters and their very results would consequently be underestimated.

Photos: 1, 2, 3, 4, 5

The aim of the lecture is to present the characteristic uses of statistical methods in psychology. We shall begin by providing brief answers to the following questions: what is psychology, why does it need statistics and to what extent did statistical methods contribute to the formation of psychology as an independent scientific discipline. In the second part we shall review the typical methods within the two main theoretic paradigms in psychology, namely experimental and correlational psychology. We shall also describe the influence statistics had on the two most important approaches to the psychological measurement. Further, we shall note some psychological / psychometric contributions to the statistical science, and some points that psychometricians have yet to learn from applied statisticians working in other disciplines. We shall conclude by a consideration of the professional identity of the psychological statisticians.

Slides: PowerPoint (in Slovenian)

Photos: 1, 2, 3, 4, 5, 6, 7

In business surveys, error reduction work has traditionally focused on sampling errors, frame errors, and nonresponse, and to a lesser extent on measurement errors although experience and sporadic research show that such errors are often nonnegligible. It is particularly challenging to find out why and how these errors occur since this is a starting point for their reduction and prevention. Therefore, the response process has to be examined in detail.

The presentation will outline the models of the response process relevant for business surveys as well as the findings of an empirical study of the response process in one of the business surveys conducted by the Statistical Office of the Republic of Slovenia.

Photos: 1, 2, 3, 4, 5, 6

DECISION SUPPORT FOR PUBLIC HEALTH PRACTICE: RESEARCH FINDINGS IN THE DEVELOPMENT OF KNOWLEDGE MANAGEMENT AND DISEASE SURVEILLANCE TOOLS
10 July, 2007
Prof. Sherrilynne Fuller, PhD
University of Washington, Seattle

The greatest improvements in the health of the public have resulted from a variety of initiatives that have had little to do with the treatment of illness - including improved water quality, food safety, injury prevention and other population health interventions. Research at the Center for Public Health Informatics is focused on the development of information systems, tools and technologies to support individuals engaged in public health research and practice. This research is currently focused in two areas - knowledge management for public health practice and design of disease surveillance systems. Knowledge management research includes studying the information needs and workflow in public health practice as well as designing content management systems with easy to use interfaces to a variety of databases, information resources and data. The second area of research that will be discussed is disease surveillance systems. This work is focused on identifying trends in disease data to: 1) provide early alerts regarding potential disease outbreaks such as avian influenza; 2) develop easy-to-use geospatial tools to support mapping of disease outbreaks or to monitor and identify environmental and other aspects of acute and chronic disease. A working prototype knowledge management system, myPublicHealth, will be demonstrated.

Slides: PowerPoint

Photos: 1, 2, 3, 4, 5, 6, 7, 8

ROUND TABLE ON THE REFORM OF THE GRADUATE STUDY IN STATISTICS

Instead of a regular lecture, we organised a round table on the reform of the graduate study in Statistics. We are obliged to reform the study because of the changes in the graduate study organisation in Slovenia (the Bologna process), but this is also an opportunity to think about possible substantial changes and ammendments of the curriculum. The round table was held on 15 June, 2007, at the IBMI. In addition to the members of the Curriculum Council and the representatives of the University of Ljubljani, it was attended by some graduate students and other statisticians. The discussion was lively and fruitful.

EVOLUTION OF NETWORKS - PROBABILISTIC INDUCTIVE CLASSES OF GRAPHS
24 May, 2007
Natasa Kejzar
University of Ljubljana, Faculty of Social Sciences

Network analysis has experienced a big breakout in the last decade in the natural sciences (development of several network generating models, analysis of processes on networks), as well as in the social sciences (analysis of structural characteristics of units and/or subsets of units in the network, application of methods to real-life data). I will present some important general characteristics of networks, main models for description of network evolution, which had large impact on the network research community, and emphasize the importance of the analyses of processes on networks.

Graphs are skeletons of networks. Many models of network evolution model only this skeleton, therefore many existing network models can be represented (rather then by stochastic processes) by inductive definition of classes of graphs, where we impose probability space for the choice of (1) basic graphs, (2) generating rules, and (3) subgraphs on which the rules can be applied.

I will present some simplified descriptions of real-life networks (e.g., spreading of rumors, changing the structure of acquaintanceships network) and basic results regarding size and order of obtained graphs, followed by a method for determining asymptotic degree distribution in such graphs.

Slides: PDF (in Slovenian)

Photos: 1, 2, 3, 4, 5, 6, 7

INFLUENCE OF ORAL HEALTH ON SYSTEMIC HEALTH
23 April, 2007
Prof. Timothy A. DeRouen, PhD
Comprehensive Center for Oral Health Research; School of Public Health & Community Medicine; School of Dentistry, University of Washington, Seattle

An overview of several years of research on the adverse effects of periodontal disease on systemic health was given.

The lecture was organised by the Department of Dental Medicine of the Faculty of Medicine at the Dental Clinic.

THE CASA PIA STUDY OF THE HEALTH EFFECTS OF DENTAL AMALGAM IN CHILDREN
20 April, 2007
Prof. Timothy A. DeRouen, PhD
Comprehensive Center for Oral Health Research; School of Public Health & Community Medicine; School of Dentistry, University of Washington, Seattle

Starting in 1997, 507 students of age 8-10 years from the Casa Pia School System in Lisbon, Portugal were enrolled into a randomized clinical trial designed to evaluate whether there were any health effects associated with treatment for dental caries using mercury-containing dental amalgam. Complete dental treatment was provided for all enrollees for seven years, with half of the children randomized to treatment with amalgam for posterior restorations, and the other half to resin composite material for posterior restorations. Urinary mercury was monitored, and neurobehavioral, neurological, and renal measures obtained annually. Results on primary endpoints, published in the Journal of the American Medical Association in 2006, were discussed.

The lecture was organised by the Department of Dental Medicine of the Faculty of Medicine at the Dental Clinic.

Photos: 1, 2, 3, 4, 5, 6, 7 (low quality; key slides shown)

DYNAMIC ANALYSIS OF RECURRENT EVENT DATA, WITH APPLICATION TO INFANT DIARRHOEA IN BRAZIL
30 March, 2007
Prof. Robin Henderson, PhD
Newcastle University, School of Mathematics and Statistics

This talk is based around the analysis of data from a study into the health impact of a sanitation programme carried out in Salvador, Brazil. The objective is to investigate risk factors associated with incidence and prevalence of diarrhoea in children aged up to 3 years old. In total 926 children were followed up at home twice a week from October 2000 to January 2002, from which daily occurrence of diarrhoea was recorded for each child being followed up. A challenging factor in analysing these data is the presence of between subject heterogeneity not explained by known risk factors, combined with significant loss of observed data through either intermittent missingness (average of 78 days per child) or dropout (21% of children). We discuss modelling strategies and show the advantages of taking an event history approach with an additive discrete time regression model.

Slides: PDF

Photos: 1, 2, 3, 4, 5

GOODNESS OF FIT IN REGRESSION MODELS FOR RELATIVE SURVIVAL
30 March, 2007
Maja Pohar Perme
University of Ljubljana, Faculty of Medicine, Institute of Biomedical Informatics
(defense of doctoral dissertation)

Prosojnice: PDF

Fotografije: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10

VISUALISATION OF MULTIDIMENSIONAL DATA USING CONVEX-HULL AND CONCORDANCE DIAGRAMS
6 March, 2007
Gaj Vidmar, MSc
University of Ljubljana, Faculty of Medicine, Institute of Biomedical Informatics
(defense of doctoral dissertation)

Dissertation: PDF (in Slovenian)

Slides: PowerPoint (in Slovenian)

Photos: 1, 2, 3, 4, 5, 6, 7, 8

(GENERALIZED) BLOCKMODELING OF VALUED NETWORKS
13 February, 2007
Ales Ziberna
University of Ljubljana, Faculty of Social Sciences

Blockmodeling is one of the methods of social network analysis. It is used for partitioning network units into clusters, called positions, and, at the same time, finding how these positions are connected to each other. The presentation will start with a quick presentation of classical and generalized blockmodeling for binary networks and their generalization to valued networks. The usage of blockmodeling will be shown through a demonstration of the small package "blockmodeling" for the R statistical language. The presented methods will be used on a social network and a food web.

Slides: PowerPoint (in Slovenian)

Examples: R code and data (ZIP archive)

R package and references: website

Photos: 1, 2, 3, 4

DEMOGRAPHY - THE STUDY OF POPULATION
18 January, 2007
Vojka Sircelj, PhD
Statistical Office of the Republic of Slovenia

It is customary to say that demography is a multidisciplinary field. But that claim is valid only for the last of the three steps of demographic research, i.e., the analysis of causes.

The value of the analysis of causes depends on the two previous steps, i.e., data collection and demographic analysis. We will discuss these two steps in relation to statistical analysis.

Slides: PowerPoint (in Slovenian)

Photos: 1, 2, 3, 4 ďťż