Giuseppe Arbia gave a talk at the spatial econometrics workshop in Paris today on spatially-dependent multiple comparison correction.
The basic idea is to use an effective sample size estimator within the BH procedure. Rank the p-values, and pick the smallest k p-values such that the kth p-value is bounded from above by k/n_eff * alpha.
As a consequence, struck me that we don't have an effective sample size code, or an APLE implementation (which would be useful for calculating n_eff).
I'm not sure how this correction would compare to the BY correction of N*sum_i^N(1/i), but I'd think that it would be less conservative/more powerful?
Giuseppe Arbia gave a talk at the spatial econometrics workshop in Paris today on spatially-dependent multiple comparison correction.
The basic idea is to use an effective sample size estimator within the BH procedure. Rank the p-values, and pick the smallest k p-values such that the kth p-value is bounded from above by k/n_eff * alpha.
As a consequence, struck me that we don't have an effective sample size code, or an APLE implementation (which would be useful for calculating n_eff).
I'm not sure how this correction would compare to the BY correction of N*sum_i^N(1/i), but I'd think that it would be less conservative/more powerful?