School of Mathematics and Statistics - Research Publications
Permanent URI for this collection
Search Results
1 - 10 of 115
-
ItemA Bayesian method for comparing and combining binary classifiers in the absence of a gold standard.(Springer Science and Business Media LLC, 2012-07-27)BACKGROUND: Many problems in bioinformatics involve classification based on features such as sequence, structure or morphology. Given multiple classifiers, two crucial questions arise: how does their performance compare, and how can they best be combined to produce a better classifier? A classifier can be evaluated in terms of sensitivity and specificity using benchmark, or gold standard, data, that is, data for which the true classification is known. However, a gold standard is not always available. Here we demonstrate that a Bayesian model for comparing medical diagnostics without a gold standard can be successfully applied in the bioinformatics domain, to genomic scale data sets. We present a new implementation, which unlike previous implementations is applicable to any number of classifiers. We apply this model, for the first time, to the problem of finding the globally optimal logical combination of classifiers. RESULTS: We compared three classifiers of protein subcellular localisation, and evaluated our estimates of sensitivity and specificity against estimates obtained using a gold standard. The method overestimated sensitivity and specificity with only a small discrepancy, and correctly ranked the classifiers. Diagnostic tests for swine flu were then compared on a small data set. Lastly, classifiers for a genome-wide association study of macular degeneration with 541094 SNPs were analysed. In all cases, run times were feasible, and results precise. The optimal logical combination of classifiers was also determined for all three data sets. Code and data are available from http://bioinformatics.monash.edu.au/downloads/. CONCLUSIONS: The examples demonstrate the methods are suitable for both small and large data sets, applicable to the wide range of bioinformatics classification problems, and robust to dependence between classifiers. In all three test cases, the globally optimal logical combination of the classifiers was found to be their union, according to three out of four ranking criteria. We propose as a general rule of thumb that the union of classifiers will be close to optimal.
-
ItemPsychosocial Well-Being and Functional Outcomes in Youth With Type 1 Diabetes 12 years After Disease Onset(AMER DIABETES ASSOC, 2010-07-01)OBJECTIVE: Type 1 diabetes in youth and community controls were compared on functional outcomes. Relationships were examined between psychosocial variables at diagnosis and functional outcome 12 years later. RESEARCH DESIGN AND METHODS: Participants were subjects with type 1 diabetes (n = 110, mean age 20.7 years, SD 4.3) and control subjects (n = 76, mean age 20.8 years, SD 4.0). The measures used included the Youth Self-Report and Young Adult Self-Report and a semi-structured interview of functional outcomes. Type 1 diabetes participants also provided information about current diabetes care and metabolic control from diagnosis. RESULTS: Type 1 diabetes participants and control subjects reported similar levels of current well-being but for the youth with type 1 diabetes, the mental health referral rates over the previous 12 years were higher by 19% and school completion rates were lower by 17%. Over one-third of clinical participants were not currently receiving specialist care and this group had higher mental health service usage in the past (61 vs. 33%) and lower current psychosocial well- being. Within the type 1 diabetes group, behavior problems, high activity, and low family cohesion at diagnosis predicted lower current well-being, but were not associated with metabolic control history. Poorer metabolic control was associated with higher mental health service usage. CONCLUSIONS: Type 1 diabetes participants report similar levels of current psychosocial well-being compared with control subjects, but higher levels of psychiatric morbidity since diagnosis and lower school completion rates. Psychiatric morbidity was associated with poor metabolic control and failure to transition to tertiary adult diabetes care.
-
ItemCurvature-constrained directional-cost paths in the plane(SPRINGER, 2012-08-01)
-
ItemExpanded Vandermonde Powers and Sum Rules for the Two-Dimensional One-Component Plasma(SPRINGER, 2012-09-01)
-
ItemGradient-Constrained Minimum Networks. III. Fixed Topology(SPRINGER/PLENUM PUBLISHERS, 2012-10-01)
-
ItemLinear Programming with Uncertain Data: Some Extensions to Robust Optimization(SPRINGER/PLENUM PUBLISHERS, 2012-11-01)
-
ItemOn the Connectivity of Visibility Graphs(SPRINGER, 2012-10-01)The visibility graph of a finite set of points in the plane has the points as vertices and an edge between two vertices if the line segment between them contains no other points. This paper establishes bounds on the edge- and vertex-connectivity of visibility graphs. Unless all its vertices are collinear, a visibility graph has diameter at most 2, and so it follows by a result of Plesn\'ik (1975) that its edge-connectivity equals its minimum degree. We strengthen the result of Plesn\'ik by showing that for any two vertices v and w in a graph of diameter 2, if deg(v) <= deg(w) then there exist deg(v) edge-disjoint vw-paths of length at most 4. Furthermore, we find that in visibility graphs every minimum edge cut is the set of edges incident to a vertex of minimum degree. For vertex-connectivity, we prove that every visibility graph with n vertices and at most l collinear vertices has connectivity at least (n-1)/(l-1), which is tight. We also prove the qualitatively stronger result that the vertex-connectivity is at least half the minimum degree. Finally, in the case that l=4 we improve this bound to two thirds of the minimum degree.
-
ItemProjective deformations of hyperbolic Coxeter 3-orbifolds(SPRINGER, 2012-08-01)
-
ItemUnit nonresponse errors in income surveys: A case study(Springer Science and Business Media LLC, 2012-10-01)
-
ItemLASSO model selection with post-processing for a genome-wide association study data set.(Springer Science and Business Media LLC, 2011-11-29)Model selection procedures for simultaneous analysis of all single-nucleotide polymorphisms in genome-wide association studies are most suitable for making full use of the data for a complex disease study. In this paper we consider a penalized regression using the LASSO procedure and show that post-processing of the penalized-regression results with subsequent stepwise selection may lead to improved identification of causal single-nucleotide polymorphisms.