I am setting up a run of feeding chambers to see if beetles prefer one genotype of plant or another. These insects are put into a petri dish ("chamber") with the plants.
Currently, I have done a trial with 12 plant seedlings, and 6 beetles, repeated x3. The feeding damage to each individual plant is quite variable (though there does seem to be a trend for genotypes, the variance is very large across individual plants). This data is also clearly not normally distributed.
Therefore, I am considering having more chambers (e.g. 6) with fewer plants (say 2 plants, 1 beetle), and considering each chamber as a replicate, opposed to each plant. My belief is this would reduce the variance and more clearly show and significant differences to feeding.
Is this assumption correct?
You should consider to increase the numbers of plants and replicates before any dataset would be robust.
Do it by step, say by step of 20 plants-replicates, and re-compute variance after each step until the variance is stabilized. Be patient...
A chamber with a plant species and beetle/s is a sub-sample, not a replicate. Your replicate is a group of chambers of the same plant species and beetle (at least 6, but 30 is better) so that you have variation within replicates. Replicate number should be at least 3. But to avoid pseudo-replication, your replicates should be spatially separated, not rotated (this is what you do to the subsamples). I hope this helps :)
- Badges
- Science topic
- Similar topics
- Biological Science
- Zoology
- Invertebrate Zoology
- Arthropods
- Insects