Textbooks on Design Of Experiments invariably start by explaining why one-factor-at-a -time (OAT) is an inferior method. Here we will show that in a model with all interactions a variant of OAT is extremely efficient, provided that we only have non-negative parameters and that there are only a few large parameters. In the extreme, this means that there is one positive parameter. In that case, for m variables, that is, for 2m-1 parameters, the procedure only needs m+2 observations to find. In other examined cases, the proposed method is also very fast.
|Place of Publication||Tilburg|
|Number of pages||14|
|Publication status||Published - 2010|
|Name||CentER Discussion Paper|
- design of experiments