I am working on modeling cyber risk and collected a sample of 500 responses. While working on them I found some of the variables do not fit into any of the known distributions. They are way off at tails. How should I proceed ?
Perhaps, it points to another character of little known interdependencies in the distribution at the larger levels of dispersion. It may happen, if the risk analyses didn't consider weaker or better hidden vulnerabilities. I'm just guessing ...