Alexander Jarman - Small-Number Statistics, Common Sense, and Profit: challenges and non-challenges for hurricane forecasting

Part of the PhD Seminar Series that alternates locations between the Grantham Institute at Imperial and the Grantham Research Institute at LSE.

Speaker: Alexander Jarman, Grantham Research Institute on Climate Change and the Environment at LSE

Abstract of seminar

When making only one forecast per year, or per decade, it can take some time to establish statistical confidence in the skill of a given forecast scheme. Must a risk-tolerant decision maker wait decades until skill is ‘proven’ if that decision maker believes the system to have value? What of a risk neutral decision maker?

A methodology was illustrated to demonstrate that there are imperfect forecast systems which almost certainly have nontrivial value long before one might establish that their skill is statistically significant.