## Important Allocation Rules

**OS = Optimal Sequential Allocation**
- defined recursively, via dynamic
programming
- evaluated via dynamic programming and backward induction
- often viewed as too computationally complex to determine

**MP = Myopic Allocation**
- one stage look-ahead rule: at each stage, determine which option
would be optimal if that were the last stage.
- easily determined directly (at each stage)
- evaluated using backward induction

**BF = Best Fixed Allocation**
- defined to be the optimal rule when one must fix the number of
observations of each population in advance
- risk and sample sizes can be obtained directly once the appropriate
equations have been derived
- may be randomized
- does not utilize information as it acrues

**EA = Equal Allocation** important special case of fixed
allocation

**HP = Hyperopic Allocation**
- adaptive version of BF rule: given current information, what is
best fixed allocation for the remainder of the experiment?
- determined directly via general solution for BF allocation (at each
stage)
- evaluated using backward induction
- may be randomized (given the best remaining allocations for each
population
using the current information, the next observation can either be
from the population needing the most allocation, or one can
select with probability proportional to the allocation needed)
- although it is a natural adaptive extension of the BF rule
it is rarely studied, and this terminology is new

Results
Outline
Previous Work