Description
Slides:
Using votes to combine rankings
Abstract:
How to best combine the outcome of several ranked data into a single, meaningful consensus ranking is a recurrent problem in scientific data analysis. Particularly, in genome data analysis we are often bound to merge ranked data arising from separate statistical analyses.
Traditional blending strategies applied in the field rely on techniques to combine statistical significance, but this approach on its own has been associated with a number of caveats.
We hereto present a voting-based heuristics implemented in Python which leverages both Schulze’s voting algorithm and optimization techniques to combine rankings upon credibility scores inferred from prior knowledge. This rationale can be used alongside state-of-the-art methods to systematically incorporate prior knowledge, thereby leading to more interpretable outcomes. The scope of the method is quite general and may be of use in other data analysis contexts.
Contributors:
Presenter: Ferran Muiños
Authors of the work: Ferran Muiños, Francisco Martínez-Jiménez
Focus:
To showcase a scientific data analysis problem arising from the study of cancer biology and how we approached it by implementing our own tool in Python.
Target Audience:
The talk aims to a broad Python audience. Minimum Python fluency is required. Acquaintance with basic notions of data analysis may be helpful to best follow the talk.
in __on venerdì 20 aprile at 16:45 **See schedule**