Efficient analysis of split-plot experimental designs using model averaging

Split-plot experimental data are often analyzed as if the came from a completely randomized design. As is well known, ignoring different levels of randomization and replication can lead to serious inferential errors. However, in some experiments, including many ocean global change experiments that motivated this research, variation between whole-plot units may be small relative subplot units. Even though factorial analysis will perform poorly general, special case it outperforms split-plot analysis, providing narrower confidence intervals for treatment means differences with coverage rates close desired level. The performance proposed model-averaged was compared classical via simulation study, its utility demonstrated an experiment examining growth condition juvenile mussel species. In our or comparisons were up 40% than while maintaining nominal rates. example experiment, we observed narrowing 25%. We recommend model averaging preferred approach when expected less units, few caveats studies very replicates.