Abstract
Justifications usually given for adopting an automated system pertain to a reduction in labor and an improvement in quality control. A manufacturer of a prototype instrument that automated some of the steps for culturing bacteria wanted to compare the automated system to the manual system. The manufacturer wanted to compare the two systems in 1) Total time needed to isolate the target bacteria, 2) Ability to isolate the target bacteria, 3) Amount of interference from background (non-target) bacterial growth, and 1) Extent of cross (sample to sample) contamination.
This paper presents the experimental design used to make these comparisons and how the design helped discover some surprising results about laboratory quality control. The experiment presented illustrates the importance of a good experimental design, the power of current statistical tools, and that a thorough and appropriate analysis of a data set requires side-by-side good detective work by both statistician and client.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Galland, J. C.; Milliken, G. A.; Hyatt, D. R.; Hornback, M.; and Cudjoe, K.
(1999).
"THE POWER OF STRUCTURED DESIGNS AND MIXED MODELS IN A REAL WORLD EXPERIMENT,"
Conference on Applied Statistics in Agriculture.
https://doi.org/10.4148/2475-7772.1271
THE POWER OF STRUCTURED DESIGNS AND MIXED MODELS IN A REAL WORLD EXPERIMENT
Justifications usually given for adopting an automated system pertain to a reduction in labor and an improvement in quality control. A manufacturer of a prototype instrument that automated some of the steps for culturing bacteria wanted to compare the automated system to the manual system. The manufacturer wanted to compare the two systems in 1) Total time needed to isolate the target bacteria, 2) Ability to isolate the target bacteria, 3) Amount of interference from background (non-target) bacterial growth, and 1) Extent of cross (sample to sample) contamination.
This paper presents the experimental design used to make these comparisons and how the design helped discover some surprising results about laboratory quality control. The experiment presented illustrates the importance of a good experimental design, the power of current statistical tools, and that a thorough and appropriate analysis of a data set requires side-by-side good detective work by both statistician and client.