Lesson29 Confounding and Pseudoreplication

Poor experimental design can lead to statistical confounding and pseudoreplication, and ultimately, to ambiguous if not misleading interpretations of results. This lesson discusses what constitutes pseudoreplication and how it's different from experiments with true replication.

[@matthiasr2739]

Thank you, this is really helpful. Well teached and explained

[@carochong1]

Than you for the video! Very helpful.

[@MrArunavadatta]

nice lecture

[@arckopolo]

This was great! It really helped me to put pseudoreplication into a more practical way of looking at it. However I do have a small question. If you are working in the field and you have a treatment that cannot be isolated to an individual (for example the fertilizer experiment you mentioned),in order to avoid pseudoreplication would you be best off increasing your study site area so you can randomly assign treatments? Cheers

[@biometryonlinelessons598]

Generally, yes. But you do what is practical, remembering that you need replication of plots within treatments (if you can't apply treatments to individuals), and that replication will determine the MS error for your treatment effect; thus the larger the N of plots, the smaller the MS error, and the larger F. In the end, it's a balancing act between practicality of treatment application, how large you want your statistical universe to be, and number of individuals you are willing to measure. Sometimes, for example with growth chambers, the experimental unit is limited by resources available (number of growth chambers). If you can't have replicate growth chambers for practical reasons, there are other strategies that can help; e.g., rotating treatments among growth chambers.