If I have categorical variables for 'floral resources used' in one column, e.g. Erica cinerea, etc. and two other columns for Pollinator Species Richness and Abundance, what test could I use? The goal is to determine which plant species are preferred by pollinators.
I tried fit regression model in Minitab, set total pollinators and species richness as responses and floral resource used as the categorical predictor. Result said floral resource used was highly significant (p < 0.001) . Coefficients section gives a list of p values for each plant species, e.g.
Erica cinerea = 0.004
Erica tetralix = 0.001
Rubus fruticosus = 0.000
Ulex europaeus = 0.802
That looks alright, but I'm not sure if I'm supposed to look at the coefficients section. Also, does that really tell you if there's a preference? It's difficult because it just says 3 out of 4 are significant but what does that really mean in this context? I mean, it syncs up fairly well with the actual numbers, for example Rubus having the highest abundance and species richness, but I'm not sure if it actually answers the preferences question.
Thank you.
When using a linear regression model, for this purpose, it is best to calculate the partial coefficients of elasticity and the standard coefficients of regression.
You should definitely look at the coefficients. Note however that their meaning strongly depends on how the categorical variable was coded, which depends from a software to another (and may be changed in some of them).
If your question is specifically « which floral species is related to maximal pollinator species richness », with the underlying assumption that there is a single species which has (much) higher richness, then you should go for specific multiple testing procedures. The exact name does not come back to my memory now, but there are some that are presented in Jason Hu's book on multiple comparison procedures.
Dear Taylor Hood
In studded population you can see more than one plant and insecte.
Firstly,
It need have the variables for 'floral resources' in one Tabl.1
and in other Tabl.2 - 'fauna resources' of the same territory.
You Tabl.3 in result can be documented by the concurrent plants for the some insects
Best regards,
MTT
The analysis depends on your data. A statistical analysis will be difficult if you have plant species 1, and species richness because you only have one replicate. If there were multiple sampling locations (multiple richness estimates for each host plant), then your options are more diverse.
A program like EstimateS will give 95% confidence intervals. If the intervals do not overlap then they are significantly different.
A favored parametric multiple comparison procedure is Tukey's test. There is a wide array of other methods. Kruskal-Wallis test is a common non-parametric approach. Many other options exist. I have see Ryan-Einot-Gabriel-Welsh multiple comparison used a few times recently.
For abundant species you might be able to run an analysis by species: Species A prefers plants 2 and 5. Plant 1 is less preferred and is never found on the other plants.
The analysis will be confounded by patch size. Species 1 covers 20 ha, while species 2 covers 2 ha, and the others combined cover 0.5 ha. Classic island biogeography problem.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Column