Reproducible Research, Uncertainty Quantification, and Verification & Validation

2014-05-03T01:22:12Z (GMT) by Lorena A. Barba
<p>Slides used with my presentation in the <strong>SIAM Uncertainty Quantification Conference 2014</strong>, Minisymposium on "The Reliability of Computational Research Findings: Reproducible Research, Uncertainty Quantification, and Verification & Validation."</p> <p>The talk used an audience response system to collect <strong>True/False</strong> or <strong>Yes/No</strong> opinions on 13 statements/questions:</p> <p>1) Computer simulations create scientific knowledge. </p> <p>2) Simulation is a method</p> <p>3) A reproducible simulation does not need to be accurate.</p> <p>4) Is there value to a Reproducible Wrong Answer?</p> <p>5) Simulation is an experiment.</p> <p>6) Is V&V a pre-requisite to Reproducibility?</p> <p>7) UQ aims to give objective confidence levels for the results of simulations.</p> <p>8) Is V&V a pre-requisite to UQ?</p> <p>9) UQ presupposes verification and informs validation.</p> <p>10) Verification should be done before validation.</p> <p>11) In verification, the actual value of the error is generally unknown.</p> <p>12) Can reproducible science be uncertain?</p> <p>13) UQ is a requirement of reproducibility.</p> <p>The slide deck was here augmented with the screenshots from the audience response.</p>