Author Information

Edzard van Santen
Mark West

Abstract

Replication and randomization and are the keys for statistically valid experiments. Both are necessary components for statistically valid experimentation. Yet it is an industry wide practicein weed science research to assign treatment in the first block of a randomized complete block design in a systematic order for reasons of convenience. We investigated this practice by comparing four randomization/analysis scenarios: (i) complete randomization in all blocks, (ii) systematic assignment of treatmentsin block 1, where the best treatment was assigned to the best plot, (iii) systematic assignment of treatmentsin block 1, where the best treatment was assigned to the worst plot,and (iv) systematic assignment of reatments in block 1 but not using it in the analysis. We created 1000 simulated datasets for three levels of experimental precision and two group sizes (t=3 and t=9). Results indicate that dropping block 1 from the analysis resulted in a loss of power, as did the best to worst assignment scenario. The best to best assignment resulted in increased power that would lead to an inflated Type I error. Differences between the drop block 1 and best to worst scenarios tended to become smaller as the experiment size increased and the experimental precision decreased. The recommendation for the practice would be (1) to follow proper randomization procedures, and (2) to add an extra block to the experiment for demonstration purposes only.

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 29th, 10:30 AM

ARMedAND DANGEROUS: THE CONSEQUENCES OF NOT RANDOMIZING THE FIRST BLOCK

Replication and randomization and are the keys for statistically valid experiments. Both are necessary components for statistically valid experimentation. Yet it is an industry wide practicein weed science research to assign treatment in the first block of a randomized complete block design in a systematic order for reasons of convenience. We investigated this practice by comparing four randomization/analysis scenarios: (i) complete randomization in all blocks, (ii) systematic assignment of treatmentsin block 1, where the best treatment was assigned to the best plot, (iii) systematic assignment of treatmentsin block 1, where the best treatment was assigned to the worst plot,and (iv) systematic assignment of reatments in block 1 but not using it in the analysis. We created 1000 simulated datasets for three levels of experimental precision and two group sizes (t=3 and t=9). Results indicate that dropping block 1 from the analysis resulted in a loss of power, as did the best to worst assignment scenario. The best to best assignment resulted in increased power that would lead to an inflated Type I error. Differences between the drop block 1 and best to worst scenarios tended to become smaller as the experiment size increased and the experimental precision decreased. The recommendation for the practice would be (1) to follow proper randomization procedures, and (2) to add an extra block to the experiment for demonstration purposes only.