I have data that describes site characteristics at the location of bird nests. The characteristics include vegetation structure and landscape physical characteristics. Data is scored as continuous, categorical, and binary. There is also one circular data measure (radial degrees 0o - 360o).
I would like to determine which of these variables would best described nest activity. Nest activity is measured as a binary metric (active or inactive). I also have nest activity for the same nests spanning over 5 years.
What models could be used? I automatically gravitated to GLM but there are probably others. And are there any models that would also cope with assessing activity over the years against the site characteristics?
Any discussion/opinion on the topic would be appreciated.
Hi Eddy,
I think you could go with any (or both) of them.
For ordination, I would recommend using PcoA. Because of the mixture of variable types you will have to use Gower distances. You could even use a stepwise procedure to determine the most important variables in the ordination.
You can also select the best GL Model using stepwise procedures, and even evaluate how accurate the selected model is in predicting your response variable.
Cheers,
Hi Nicholas
Thank you for your comments. I like the idea of PCOA and the Gower distance. However, I'm still not convinced that any of the distance measures deal with circular data all that well.
Cheers
Maybe you can transform that variable according to any of the ways suggested at https://www.researchgate.net/post/What_methods_are_there_to_transform_aspect_degrees_to_avoid_the_issue_of_circular_data
Best
Thanks again Nicolas. I will stick with the GLM and compare models after stepwise addition and removal of factors. But I may use ordination (MDS) to see if any confounding variables exist (I prefer it to straight correlation).
- Badges
- Science topic
- Similar topics
- Force