We don't know. The confidence intervals on individual probability distributions are already wide and long-tailed but convolving them only makes the problem wors

We don't know. The confidence intervals on individual probability distributions are already wide and long-tailed but convolving them only makes the problem worse. https://arxiv.org/pdf/1806.02404.pdf https://arxiv.org/pdf/1806.02404.pdf