“the use of inferential statistics to test for treatment effects with data from experiments where either treatments are not replicated (though samples may be) or replicates are not statistically independent” (Hurlbert 1984)
Hurlbert, Stuart H. 1984. “Pseudoreplication and the Design of Ecological Field Experiments.” Ecological Monographs 54 (2): 187–211. https://doi.org/10.2307/1942661.
Replication is repeating an experiment.
Pseudoreplication happens when treatments are not actually replicated or replicates are not
Here is a simple youtube video on Pseudoreplication, and another on Biological & Technical Replicates by Henrik’s Lab
Technical replicates increase the accuracy of the same sample, whereas biological replicates are measurements from different samples and capture the biological variation.
Basic tents of experimental design:
experimental variables within a treatment must not influence each other more than they influence the variables within another treatment.
factors other than the treatments must, on average, be equal across all treatments
(e.g. seawater source, light, salinity, etc.) must all be the same in each treatment and controlled across the experiment so that those factors do not confound the differences you see between treatment groups
Degree of precision that the treatment is applies and its effects measured
“If treatment effects are confused with the effects of other factors not under investigation, then an accurate assessment of the effects of the treatment cannot be made.”-Cornwall and Hurd (2016)
Simple Pseudoreplication
One ‘experimental unit’ per treatment and multiple individuals in that treatment whose response is being measured.
Temporal Pseudoreplication
Multiple measurements are made through time on the same experimental unit and treated as independent.
Sacrificial Pseudoreplication
Multiple ‘experimental units’ per treatment and multiple individuals within each experimental unit, but the individuals within each experimental unit are treated as experimental units during the statistical analysis.
“These three definitions demonstrate how a misinterpretation as to what constitutes an experimental unit vs. an”evaluation unit” could lead to inappropriate design and analysis.” ((Cornwall and Hurd 2016), p. 573)
Cornwall, Christopher E., and Catriona L. Hurd. 2016. “Experimental Design in Ocean Acidification Research: Problems and Solutions.” ICES Journal of Marine Science 73 (3): 572–81. https://doi.org/10.1093/icesjms/fsv118.
Implicit Pseudoreplication
Experimental tank arrays:
Tank types:
Storage tank: seawater stored before altered
Mixing tank: where chemicals and seawater are mixed to create treatments before contact with the study species
Header tank: treatments stored before exposure to study species
Experimental tank: the ‘experimental unit’ where study species is housed in the treatment
Randomized Complete Block Design
blocks are constructed around known or suspected source of variation
each block contains all treatments
blocks are relatively uniform
puts experimental units that are as similar as possible together in the same block, and assign all treatments into each block separately and independently
variation among blocks can be measured and removed as experimental error
Factorial Design
tests for interaction between multiple variables
- (thermal stress & pollution stress)
all treatments of one factor should be tried with all treatments of the other factors
Further Literature on Pseudoreplication: