I would be tempted to use a simple cluster analysis first.

How many isotopes do you have for your analysis? What other factors do you know about the dingo and their prey? 

Thanks Andrew

I don't have very big sample sizes for my prey-species isotope data, which is why I wanted to use KNN randomization. I managed to get my hands on some code to run the analysis today, and as I suspected, many of the species were not significantly different (there's a lot of variability).

I have categorised the isotope data using cluster analysis, as part of another aspect of my work in which I'm looking at habitat use. Fortunately for me there are some nice, discreet clusters when I don't have to worry about the species.

Thanks again

In a data set I worked with, we had data on the mercury content of fish from lakes and a lot of info about the lake mineral content. We were able to "standardize" the Hg content of the fish based upon their age. Older and heavier fish have more mercury the younger and smaller fish have less. Once we made those calcs, we got some good cluster analysis. Similar lakes were near each other and down wind of power plants. I was wondering if you would have the same kind of luck we did. 

Hi Damian. I'm trying to test for differences among groups of prey as well and see that knn is one method people are using. But through reading a little bit online, I don't quite understand how it tells you whether your groups are different or not--I thought its main purpose was assigning new values to groups based you training it on what belongs in what group...Might you be able to pass along any info that made you understand how to actually do/interpret a test for differences (I'm using R in case that's relevant)? In my data, for example, I'd be trying to determine if deer vs elk vs moose vs caribou, etc have different CN signatures. You can email me at [email protected] if you have any advice. Thanks much.

Hi Joy

I sent you an email

Cheers

Hello Damian and/or Joy,

I have the same question: What package/implementation did you use in R to set up this test (testing for Euclidean separation between prey source data in d15N/d13C isospace)? If one of you could post an answer or email me at [email protected], it would be greatly appreciated. Thank you.

Hello Damian and/or Joy and/or Douglas,

I too am curious about the R package and implementation used to test for separation in isotopic source data. A search online seems to yield results for kNN in contexts other than isotopic ecology. I would greatly appreciate more information. My email is [email protected]

Kind regards,

Jordon Munizzi

Damian Stuart Morrant Hi Damian,

I realized I'm a little late to the game, but I'm wondering if you would be so kind to share your kNN test scripts with me ([email protected])? Also I have a question (excuse me if it's a stupid one), if this is an one off study, what is the advantage of using a machine learning algorithm when i won't be adding more data point in the future. Thanks in advance for your help!

Best,

Chris

Hi,

I'm also very interested in trying this approach to determine source separation. If anyone is willing to share their methods and perhaps even R-scripts, please shoot me an e-mail at [email protected].

Best

/Olof

Jumping on the bandwagon! Hello Damian and/or Joy and/or Douglas--I've been trying to figure out an R package and/or code to use for kNN to help combine prey sources for running a stable isotope mixing model. If you have any resources or R scripts that you would be willing to share, could you email me at [email protected]? Thanks!

-Wieteke

I am going to jump on this bandwagon, as well! I am trying to figure out how to run k nearest neighbor randomization tests in R to determine if sources (e.g. prey types) are statistically different in dual isotopic space. Similar to what Joy Erlenbach said, I figured out how to use the knn() function in the 'class' package to determine the percent accuracy of knn prediction when assigning test observations to groups. But, what I'm really looking for is a way to statistically test for a difference between groups using knn. If anyone could send along any relevant resources or R scripts I would very much appreciate it! You can reach me at [email protected]. Thanks!