I performed a multiple choice experiment to assess the food preference of the sea urchin Paracentrotus lividus in relation to 5 different food items. I have 15 replicates, each represented by a tank where the 5 food items were simultaneously offered to 1 sea urchin. For each tank, I measured the consumption of each food item after 48h.
Given the lack of independence within each replicate (the consumption of one food depends on the presence of other food sources), I would use a multivariate approach. The problem is that no treatments have been included in my experimental design (all replicates have the same condition) and it makes difficult choosing a proper statistical approach. Any suggestions?
Thanks
I'm guessing that given the fact you're studying an urchin, consumption was measured in grams consumed or a similar continuous metric and not simply marked as eaten/not eaten. If I'm right, then the best way to handle it is to normalize the amount of each item eaten in each trial by the total amount consumed. That is calculate the percent consumption of each food type. You are then controlling for quantity consumed in each trial and accounting for that necessary dependence. You can then average across the replicates and get your answer.
Ideally for this analysis the urchin would always have had some of each food available, because if the urchin eats all of item A then that item is no longer a choice. If that is not the case or if that did not occur most of the time, then you needn't take that into account. But if that happened frequently you may need to rethink the analysis.
Thank you both for your reply. I think I've found a solution about the analysis of my data. I'm going to use a non parametric test (Friedman's test) to evaluate any differences in the amount of food eaten by the urchin within each replicate, taking into account that my observations are not indipendent.There are also tests for multiple pairwise comparisons (e.g. Wilconxon test) to run afterwards.
Many thanks
in order to show the difference between choices in a multiple choice question you would be sensible to use the one way chi squared test. This is the most conservative and sensible choice especially if you are not really genned up on statistics. To then doemonstrate the causality for this is not a matter of statistics but of having a valid hypothesis coming from some theory that you can show is chosen most frequently across the range of questions.
What is meant by Double loop and Single loop? - ResearchGate. Available from: https://www.researchgate.net/post/what_is_meant_by_Double_loop_and_Single_loop [accessed Feb 15, 2016].
- Badges
- Science topic