A new measure of explained variation is defined in the paper. The measure has required statistical properties and can be used with any model and any kind of data (e.g. time dependecy, repeated events ...)
COBISS.SI-ID: 27655129
Pseudo-observations present a generalized approach to fitting various survival models and checking their assumptions. The advantages of the method have been studied in various situations.
COBISS.SI-ID: 26578137
Standard classification methods used on class-imbalanced data often produce classifiers that do not accurately predict the minority class; the prediction is biased towards the majority class. In this paper we evaluate if the high-dimensionality of data poses additional challenges when dealing with class-imbalanced prediction. We identify the additional sources of bias specific of the high-dimensional data, and investigate the effectiveness of some strategies that are available to overcome the effect of class imbalance.
COBISS.SI-ID: 27503321
We describe a method for categorization of MEDLINE citations. The motivation for themethod is the gene symbol disambiguation problem, which is very important for information extraction from biomedical texts. The basis of the method is a Chi-square-based scoring function, which is calculated for each and every MEDLINE citation. With this method we categorize bibliographic citations into genetic and non-genetic category and then we extract gene symbols only from the genetic category of citations. The method is very fast and it can categorize the whole MEDLINE database in a few hours.
COBISS.SI-ID: 26772441
This manuscript presents and evaluates an objective approach to test and select an appropriate AT for computer access for people with disabilities. Six user interfaces were tested on the control group of 29 people without disabilities and on 63 people with neuromuscular diseases, using purpose-built software for testing the speed of sentence typing. Different criteria for selecting the optimal AT were tested and compared with the skilled clinician’s choice. The learning curves of the people with disabilities proved to follow those of the healthy controls, but with lower performance.
COBISS.SI-ID: 1051753