Projects / Programmes
Methodology for data analysis in medical sciences
January 1, 2015
- December 31, 2021
| Code |
Science |
Field |
Subfield |
| 3.08.00 |
Medical sciences |
Public health (occupational safety) |
|
| 1.01.00 |
Natural sciences and mathematics |
Mathematics |
|
| Code |
Science |
Field |
| B110 |
Biomedical sciences |
Bioinformatics, medical informatics, biomathematics biometrics |
| Code |
Science |
Field |
| 3.03 |
Medical and Health Sciences |
Health sciences |
biostatistics, survival analysis, relative survival, explained variation, high-dimensional data, classification, class-imbalance, boosting methods, regression modellingquality in health-care, bibliometry, literature-based discovery.
Researchers (33)
Organisations (1)
Abstract
In biostatistics our research focuses on three areas:
· Explained variation in survival analysis has the longest tradition at our department and we have published several papers on the subject. In the near future we intend to work on the required properties of these measures, and we believe that the present understanding of these properties, bias in particular, is wrong and should be changed. The focus should then turn to other properties, of which comparability across models and understanding of the scale seem to be the most important.
We plan to extend the definition of our Re measure to competing risks and multistate models, and we are also collaborating on a new measure.
· In the area of relative survival we intend to further contribute to the methods of analysis, to which we have considerably contributed in the past. The main focus will be on our recently introduced estimator of the net survival which proved to be the only consistent method for the purpose. We will study its behaviour and the necessary adjustments in case of informative censoring or discretely collected data and provide software to enable wide usage of all the methodology. We believe the pseudo-observations approach could be helpful in solving the above-mentioned problems.
· We will also continue to work in the fast-growing area of high-dimensional data, where the number of variables greatly exceeds the number of observations. Our research focuses mainly on classification problems for class-imbalanced data and on the use of specific regression modelling techniques. We plan to continue our on-going assessment of the performance of the existing methods in these settings, expanding our studies to multi-class problems. We will focus also on boosting methods and penalized logistic regression. Our preliminary results showed that the performance of both methods on high-dimensional (and class-imbalanced) data might be greatly improved. We wish to identify the reasons for their poor behaviour and to propose solutions that might improve their performance. We plan to expand our research focus also to intensive longitudinal data, where variables are repeatedly measured over time and the number of repeated measures is large.
In the area of literature based discovery we plan to work in two directions:
· Further development of our tools to better deal with graphs, including whole exome sequencing and testing the tools of the Semantic Web.
· Practical applications, like studying adverse drug effects and using nutrigenomic data to suggest appropriate nutrition for specific diseases.
In bibliometry we will investigate different measures of scientific quality, with special emphasis on the number of pure citations per author, which has long been used in Slovenia in the field of medicine.
In the area of health care quality monitoring we intend to continue work on our approach to outlier detection and to co-operate with medical researchers to possibly define new quality indicators for certain subfields of medicine.
Significance for science
One can easily overstate the importance of his/her work for the development of science. We are certain that in the fields of biostatistics, bioinformatics and bibliometry we have done a lot in the past years, and that this is evident from our publication record. We are especially proud that our work has been well received and cited, which is reflected in the citations and invitations for talks at various international conferences and institutions. Data on citations are evident from the bibliometric analysis of our group, here I would like to stress that our paper in Biometrics in 2012 has been the most cited paper of this journal among the papers published in 2012 and 2013. Here is the link to this information http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1541-0420/homepage/MostCited.html. Biometrics is one of the most prominent journals in th area of biostatistics.
We have also published papers in top journals in other areas, most recently two papers in the Journal of Informetrics, which occupies the second place among the 84 in its area.
Our methodological research has direct effect on our support of research in Slovenian medicine, where in cooperation with medical researchers we publish a few dozen papers each years. This is a lot considering the size of our group.
Significance for the country
Codes at the end of each paragraph are used by the Research Agency to classify socio-economically important activities.
The members of the programme group are lecturing at the undergraduate and graduate level at 4 Faculties of the University of Ljubljana (Fac. of Medicine – Medicine and Dental Medicine), Veterinary Fac., Fac. of Health Sciences – 4 departments, Fac. of Electrical Engineering – MSc in Applied Statistics) in 13 courses (9 from the field of statistics, 4 from informatics); at 2 Faculties of the University of Primorska (Fac. of Mathematics, Sciences & IT, Fac. of Health Sciences) in 3 courses; and in 3 courses of the interdisciplinary PhD programs in Biomedicine and Statistics at the University of Ljubljana. (B2 2.2)
During the last 10 years, members of the group have served as advisors for 5 PhD's and co-advisors for 4 PHD's, advisor for 1 co-advisor for 1 pre-Bologna Master degree, co-advisors for 2 MSc's, 2 times as advisors and 7 times as co-advisors for pre-Bologna Bachelor degrees, and once as advisor and 4 times as co-advisors for BSc's. (B2 2.2)
Maja Pohar Perme published a newspaper article in the Science supplement of the Delo daily entitled "Statistics fighting doping: biological passport is highly likely to proclaim some innocent as guilty" (http://m.delo.si/clanek/155859) (B2 2.4)
On 15 November 2012, members of the group participated in the national TV show Good Hour dedicated to statistics (http://ava.rtvslo.si/predvajaj/dobra-ura/ava2.150593290/) (B2 2.4)
Janez Stare serves as Chairman of Program Council of the University Doctoral Study in Statistics, and as Vice-Chairman of the Program Council of the Master Study in Applied Statistics. (B2 2.6)
Gaj Vidmar has been serving as Editor-in-Chief of the Informatica Medica Slovenica journal since 2009 (preceded by Janez Stare) and as Associate Editor of the International Journal of Rehabilitation Research (Impact Factor above 1). (B2 2.6)
Maja Pohar Perme is a member of the Program Council of the Master Study in Applied Statistics. (B2 2.6)
Lara Lusa serves as Editor of the journal Advances in Methodology and Statistics. (B2 2.6)
Janez Stare has been chairing the International Program Committee (IPC) of the Applied Statistics international conference; Gaj Vidmar and Lara Lusa have served as IPC members. (B2 2.6)
Maja Pohar Perme serves as Editor of the Slovenian Medical Journal (Impact Factor above 0.1). (B2 2.6)
Janez Stare was a member of the Program Committee of the ISCB Annual Conference in Ottawa, 2013; Maja Pohar Perme was a member of the Program Committee of the ISCB Annual Conference in Munich, 2013. (B2 2.6)
Janez Stare serves as a member of the Statistical Council of the Statistical Office of the Republic of Slovenia (SORS). (B2 2.6)
Rok Blagus is a member of the Methodological Council of the SORS. (B2 2.6)
We performed 5-10 large-scale data analyses yearly for pharmaceutical companies. (B2 2.7)
We have developed a large amount of software for the R environment; the relsurv package is frequently used abroad and in the Cancer Registry of the Republic of Slovenia. (B2 2.8)
We have developed and we are maintaining various information management systems (secure and reliable multicentric data collection and storage with a complex search engine and the option of publishing statistical analyses) for clinical research, registries and quality management (for the National Institute of Oncology, pharmaceutical companies, health care providers etc.). These systems are built on our own platform, which is characterised by extreme adaptability to the data types (data models) and end-user- friendliness. (B2 2.8)
Most important scientific results
Annual report
2015,
interim report
Most important socioeconomically and culturally relevant results
Annual report
2015,
interim report
