Chris Smeenk (Western University), Sarah Gallagher (Western University)
Cosmologists have pursued a variety of strategies to assess the reliability of N-body simulations. We will focus on the viability of robustness analysis (Weisberg 2006). We argue that robustness analysis alone is not sufficient to establish reliability (following Parker 2011). A set of simulations may robustly share certain qualitative features because they have all been "tweaked" to yield the same results, rather than due to some underlying common structure. We will consider the conditions that would have to obtain over an ensemble of models to counter this worry, and argue that these conditions would be difficult to achieve.
