Title :
Evaluating Statistical Tests on OLAP Cubes to Compare Degree of Disease
Author :
Ordonez, Carlos ; Chen, Zhibo
Author_Institution :
Dept. of Comput. Sci., Univ. of Houston, Houston, TX, USA
Abstract :
Statistical tests represent an important technique used to formulate and validate hypotheses on a dataset. They are particularly useful in the medical domain, where hypotheses link disease with medical measurements, risk factors, and treatment. In this paper, we propose to compute parametric statistical tests treating patient records as elements in a multidimensional cube. We introduce a technique that combines dimension lattice traversal and statistical tests to discover significant differences in the degree of disease within pairs of patient groups. In order to understand a cause--effect relationship, we focus on patient group pairs differing in one dimension. We introduce several optimizations to prune the search space, to discover significant group pairs, and to summarize results. We present experiments showing important medical findings and evaluating scalability with medical datasets.
Keywords :
data mining; diseases; medical computing; patient treatment; statistical testing; OLAP cubes; diseases; lattice traversal; medical domain; medical measurement; medical treatment; multidimensional cube; parametric statistical test; patient records; risk factors; statistical tests; Disease prediction; OLAP; disease prediction; parametric test; Algorithms; Diagnosis; Disease; Humans; Models, Statistical; Predictive Value of Tests;
Journal_Title :
Information Technology in Biomedicine, IEEE Transactions on
DOI :
10.1109/TITB.2008.926989