2
$\begingroup$

I am describing a new species of ant. Molecular and Morphological data suggests that there are 2 separate species. In fact we are very certain they are separate based on behavioral information. I would like to statistically analyze the measurements I have taken from my samples to support what I have found morphologically and molecularly.

However, I am having trouble with where to begin with my statistical analysis, or what sort of analysis I should do.

Issue: I have a data set of n=42 Colonies, with n=172 individuals measured; 9 measurements and 5 indices for each worker (indices are standard practice). Broken down further morphologically; I have n=13 colonies and n=46 workers from the described species; and n=29 colonies and n=127 workers from the new species. The biggest issue is the sampling error for the measurements. This is from how uncommon the described species is, and its confusion with the undescribed species. I have some colonies sampled that had 2-3 workers measured, while most had 5 workers measured, a few with 5+ workers measured , and even some that were measured as the described species but should have been measured as the undescribed. We did not have behavioral information for those colonies that were misplaced.

I have two questions I want to answer: Are the measurements from each colony are similar (Variance in a colony)? Show that they are separate species from these measurements. When I measured total length of the ants, it strongly suggested they are separate species.

My preliminary data showed that they were separate, but I only used a Z-test. Here are two characters we reported on a poster. A z-test was "helpful" but not really informative or useful without an ANOVA, which I can't do since the measurements are dependent of the colony and the sampling issue. Data from my research poster

I had started to do a principal components analysis of the measurements (using PNG) with 4 principles resolving ~90-95% of the variance; as suggested by a colleague. The scatter plot did show the expected grouping by species(Red= Described, Grey- Undescribed, Yellow= Should have been measured as undescribed). But I am not certain of what to do with that data or if it really resolves my issue of dependent variables. PCA analysis of all 14 measurements

I had considered finding the averages of colonies, since each worker sampled came from the same mother and has the same genetic information as its sisters. But the sampling issue arises. Are the measurements I took of workers from a colony replication for that colony? Should the type specimen of the described species I measured hold more weight than other specimen of that species I measured (I measured a paratype and 3 holotypes)?

Please let me know if I should provide more information. Or there is something wrong with my train of thought/reasoning. Thank you.

$\endgroup$
  • $\begingroup$ Welcome to Cross Validated! You'd be better off dropping the first paragraph, in which apologies & excuses for a lengthy question lengthen it still more. Clarifications you might edit the q. to provide include: (1) What's a measurement & what's an index? (2) Why is there one more worker than there are individuals? (3) By "Sampling error" you seem to mean mis-identification of colonies. Is that so? And why are you telling us if you've identified them all right now? (4) What's PNG stand for? (5) What are "types", "paratypes", & "holotypes"? $\endgroup$ – Scortchi Jan 6 '17 at 20:56
  • $\begingroup$ You may apply multivariate cluster analysis. If you finish with two clusters where one of them will contain predominantly described species and the other one - undescribed, it will strongly support your hypothesis concerning new species. Also if you will use just 2 factors for clasterization it could be beautifully presented with a picture where different species will be spaced differently. $\endgroup$ – Bogdan Jan 6 '17 at 21:02
  • $\begingroup$ Thanks @Scortchi. 1) A measurement is say the length or width of the head, the head index (Cephalic index) = 100*(Head width/Head Length). 2) I think That was an error on my part, individuals=workers. 3) Sampling error probably refers to the sample size of workers measured from a colony, and uneven sampling. I was limited by the Number of workers from the described species, and not the undescribed species. 4)PNG is a statistical analysis software from SAS. 5)Holotype = The specimen the species are described from, Paratypes support the Holotype specimen description. $\endgroup$ – nepenthesbaphomet Jan 7 '17 at 5:38
  • $\begingroup$ @Bogdan I will look into the Multivariate cluster analysis. I appreciate your input! $\endgroup$ – nepenthesbaphomet Jan 7 '17 at 5:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.