kaitlyngaynor
commented
3 years ago The first was about whether there's any reason not to include all data that we have (so our question about whether we need to restrict to only cameras that have sufficient data across all 4 years). Only one person has responded, and his answer was no, there's no reason not to use everything we have "Unless the variability of camera operation has something to do with characteristics of the site" (https://groups.google.com/g/unmarked/c/l2aWnibSDW4). My understanding of the cameras is that we should be fine--what do you think?
I would think that we are fine. I pulled the cameras that are furthest from roads, but given that we aren't testing changes in the effects of roads, I don't think that it will matter. I might also advise running analyses with and without these cameras just to see if they indeed change responses in a meaningful way.
Again, I think that means we're fine continuing as planned? I'm pretty sure I got a full detection history created the other day using the new camera operation spreadsheet from you and Meredith, so my next step will be getting it into the model and figuring out how to make sure the variables we're interested in are also in the model.
I'm not sure I 100% follow, but I'm assuming that you are somehow controlling for the session/year later in the modeling already? I think you should continue as planned.
I asked two questions in the google group:
The first was about whether there's any reason not to include all data that we have (so our question about whether we need to restrict to only cameras that have sufficient data across all 4 years). Only one person has responded, and his answer was no, there's no reason not to use everything we have "Unless the variability of camera operation has something to do with characteristics of the site" (https://groups.google.com/g/unmarked/c/l2aWnibSDW4). My understanding of the cameras is that we should be fine--what do you think?
I also asked whether I needed to use the "sessionCol" argument for creating my camera operation matrix to feed into a umf for a multi-season model, and got this response (I'd asked if you'd run into issues by having a single row for every camera with the down times as problems instead of using a labeled session):
it depends what you mean by issues. Technically you can do it. What is more relevant is the assumptions that are implied, most importantly the closure assumption. It will not hold over many years. Also, using multi-sessions allows you to estimate additional parameters (extinction and colonization probabilities) and how they are affected by covariates. The best choice depends on the study design and how it relates to the assumptions of the models. See the colext vignette in unmarked for some thorough explanations and examples.
Again, I think that means we're fine continuing as planned? I'm pretty sure I got a full detection history created the other day using the new camera operation spreadsheet from you and Meredith, so my next step will be getting it into the model and figuring out how to make sure the variables we're interested in are also in the model. Hope you had a nice weekend!