Skip to main content

D. Douglas' Manual Classification Tools: Segmentation, methods versus objectives, D. Douglas 21 Aug '09

Greetings, Roland's post last week got me thinking about segmentation (again), and prompted me to share some thoughts that stem from an ad-hoc segmentation method that I've been developing and tinkering with. Please see the attached/accompanying PDF file. Best regards, Dave

R code for behavioural change point analysis

I have taken a look at the paper and the R code. I must say the code does not really seem complete and is certainly far from self explanatory. I downloaded the code from the journal website. In short, no success either.

I will try to make the time to look through Dave's document.


velocity is better than displacement

David, thanks for sharing your threshold-based algorithm. You are right that the squared net displacement is limited to animals with fairly simple out/back migrations. It would probably not be useful for more circuitous routes, and I'm not sure how well it would highlight stop-overs. The advantage is that its easy to run in R and helps view these drastic changes in behavior.

I like your tool because it lets users set thresholds, see results, modify thresholds, until they have something that accurately represents the shades of grey (vs. black/white) of migratory behavior including stopovers, migrations, and 'wandering'. These thresholds are also simple and easy to report in the methods of a paper.

What are the practicalities of running your algorithm? What program? Does it output the results with the location file, or would you go in and manually classify points based on looking at this graph?

By focusing on turning angles and velocity to classify behavior your tool is similar to some of what Ferdi and Fred presented at Hawk Mountain. Also similar to the recently published behavioural change point analysis (BCPA)

I'm still wrapping my head around some of the BCPA methods, I have posted the R code but not exactly figured out how to make use of it. Anyone else have more luck?