Formatting capture history

[apryle]

I surveyed for raccoons in 2 neighborhoods. Each neighborhood had 3 streams (6 streams total). I am interested in estimating raccoon abundance at each stream, and also for each neighborhood. I expect detection probability to vary by sex. I have a few questions...

1. Please confirm that this is correct... I can manually create combinations of sex-stream (i.e. male-stream1, female-stream1, male-stream2, female-stream2, etc) to use as a group term, resulting in 12 groups (6 streams x 2 groups at each stream) with an encounter history for 4 occasions that looks something like this for each raccoon...
1000 1 0 0 0 0 0 0 0 0 0 0 0 (male-stream1 encountered at occ 1)
1010 0 1 0 0 0 0 0 0 0 0 0 0 (female-stream1 encountered at occ 1 and 3)

2. Can I simply add the abundance estimates of the 3 streams to get the neighborhood abundance estimate?

3. Future datasets will include additional streams and an additional groupings for habitat type, weather, and moon phase. Is there a better way to approach this so I do not have to create all of these combinations manually? I am open to using MARK or RMark, if that would be more concise.

Thanks!

[Bryan Hamilton]

You are on the right track. Have you read the MARK book? Chapter 2 has a very thorough explanation on formatting the encounter histories and the .inp file. Your data structure seems to be nested, streams are nested in neighborhoods.

1. Depending on how many individuals you captured, 12 might be too many groups.

2. What kind of model are you using? I'm guessing a closed capture or POPAN if you're trying to estimate abundance. Have you carefully considered the assumptions of those models?

3. Unless you are already proficient in Program R, I would start with MARK. RMark does all the modeling through MARK. The learning curve for RMark is very steep, even if you understand what MARK is doing in the background.

There are lots of tutorials out there for both MARK and RMark. They helped me TREMENDOUSLY.

Best of luck.

[apryle]

Yes, I have read chapter 2 of the book. Unfortunately, I did not find it to be very thorough at all regarding multiple groups. Section 2.1.1 describes groups within groups, but does not offer an alternative to manually creating all of the combinations. The book does not provide any examples where there it is more than a 2x2 design. Furthermore, I was also hoping to stick with each line as an individual rather than combining frequency counts, as this seems like a recipe for error and constant struggle to reupdate the inp file because my data are not entered in that format.

1. I understand that 12 groups may be too many. Assuming it is not, is there a way to do this without manually creating all of the combinations. Furthermore, I'm interested in learning MARK, not just finding this one specific solution. So, I guess my question, "Can this be done without manually creating every combination" is somewhat theoretical.

2. I am using a robust Huggins model and have considered assumptions, but I'm really interested in seeking help on formatting my data as described.

Any additional help is greatly appreciated!

[cooch]

Section 2.1.1 describes groups within groups, but does not offer an alternative to manually creating all of the combinations.

Because there aren't alternatives. The .inp file is created using some facility on your part with software designed to generate the encounter histories, and frequencies of same subdivided over as many groups and categories as you want. The presumption is you have such programming facility. Failing that, you'll need to find/bribe someone who does.

The book does not provide any examples where there it is more than a 2x2 design.

Which is sufficient (which is why there is only the one example) -- if you understand the (2x2) design, then you should have no trouble extending that to any straight factorial design.

Furthermore, I was also hoping to stick with each line as an individual rather than combining frequency counts, as this seems like a recipe for error and constant struggle to reupdate the inp file because my data are not entered in that format.

...which changes nothing. So, instead of say

Code: Select all

11011 3 4;


you simply have

Code: Select all

11011 1 0;
11011 1 0;
11011 1 0;
11011 0 1;
11011 0 1;
11011 0 1;
11011 0 1;


1. I understand that 12 groups may be too many.

Too many for MARK? No. Too many to be useful? Depends on the purpose.

Assuming it is not, is there a way to do this without manually creating all of the combinations.

If by 'manual' (and I'm guessing here, since you never actually define what you mean) you mean 'having to create the frequency columns yourself in the .inp file, external to MARK, then..' -- see first answer. MARK has no facility for 'massaging your data'. That's up to you. You can either use Excel (as per Addendum to the chapter), for many problems, or R, or SAS -- or any host of applications.

[apryle]

Thank you for your reply - this answered all my questions.