You are here

Are we expressing uncertainty of our claims enough? Towards automatic detection of overstatement of claims.

Printer-friendly versionPrinter-friendly version
Amini S, Kilicoglu H, Hooft L, ter Riet G
2015. World Conference on Research Integrity (WCRI).
Abstract: 

We data-extracted 100 randomized trials from a sample of recent Cochrane Collaboration reviews. For us this was efficient because these trials had already been assessed for their methodological robustness through Cochrane’s ‘risk of bias’ scale. We extracted outcome data on main results of all trials and their statistical precision (via its standard error). Then, dedicated software determined hedging scores automatically. [2] These hedging score were normalized to account for a paper’s word-count. We determined 10th
and 90th centile normalized hedging scores and related these to (1 minus) the risk-of-bias scores (study quality).

Amini S, Kilicoglu H, Hooft L, ter Riet G. Are we expressing uncertainty of our claims enough? Towards automatic detection of overstatement of claims. 2015. World Conference on Research Integrity (WCRI).