Communications of the ACM
Computer Algorithms Can Test the Dodginess of Published Results
Researchers at the University of Illinois at Urbana-Champaign researchers have developed a method for determining whether published scientific results have been manipulated.
Credit: Getty Images
University of Illinois at Urbana-Champaign researchers have developed a method for algorithmically gauging manipulation of published scientific results, given the mean, standard deviation, and number of data points in a result (all three of which are usually stated as part of such a result).
The Complete Recovery of Values in Diophantine Systems (CORVIDS) algorithm is designed to deliver a baseline rating of the apparent plausibility of results if it can rebuild a valid dataset or sets.
CORVIDS is tasked with finding all possible combinations of numbers that address the linear equations from which the statistics being analyzed are calculated.
To more easily detect anomalies, CORVIDS converts the possible datasets into histograms and configures them along three dimensions. Missing values in the dataset at one end of the scale would imply either incorrect reporting of statistics or problems with the underlying data, which could signal biased collection techniques or even fabrication.
From The Economist
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found