3-year CJS analysis with no 101 encounter history?

questions concerning analysis/theory using program MARK

3-year CJS analysis with no 101 encounter history?

Postby jmwiniar » Fri Jul 15, 2016 1:37 pm

Greetings - I was asked to analyze a 3-year capture-resight data set for a songbird species to obtain apparent annual survival. Birds were initially color-banded in 2014, with resighting surveys and additional banding in 2015. Due to logistical restraints, there was no banding done in 2016, but resighting was conducted in 2016 for birds marked in 2014-2015.

Upon inspecting the data, I noticed there are no 101 encounter histories for any of the individuals. This will be problematic for estimating the recapture probability, correct? And I would only be able to run models where p~1? I'm new to doing this sort of analysis, and have not come across examples with similar data (i.e., lacking 101 histories)

The encounter histories (ch) for each individual bird in the data set look like this:

Code: Select all
    ch      freq
1  100    1
2  100    1
3  100    1
4  100    1
5  100    1
6  100    1
7  100    1
8  100    1
9  100    1
10 100    1
11 100    1
12 100    1
13 100    1
14 100    1
15 100    1
16 100    1
17 100    1
18 100    1
19 100    1
20 100    1
21 100    1
22 100    1
23 100    1
24 100    1
25 100    1
26 100    1
27 100    1
28 100    1
29 100    1
30 100    1
31 110    1
32 110    1
33 110    1
34 110    1
35 110    1
36 110    1
37 110    1
38 110    1
39 110    1
40 111    1
41 110    1
42 110    1
43 110    1
44 111    1
45 111    1
46 110    1
47 110    1
48 110    1
49 110    1
50 110    1
51 110    1
52 110    1
53 010    1
54 010    1
55 011    1
56 010    1
57 010    1
58 010    1
59 010    1
60 011    1
61 010    1
62 010    1
63 010    1
64 010    1
65 010    1
66 011    1
67 010    1
68 011    1
69 010    1
70 010    1
71 010    1
72 010    1
73 010    1
74 010    1
75 010    1
76 011    1
77 010    1
78 010    1
79 010    1
80 010    1
81 011    1
82 011    1
83 010    1
84 010    1
85 010    1
86 010    1
87 010    1
88 010    1
89 010    1
90 010    1
91 010    1


Thanks,

Jason
jmwiniar
 
Posts: 3
Joined: Mon Jul 11, 2016 10:34 am

Re: 3-year CJS analysis with no 101 encounter history?

Postby Morten Frederiksen » Mon Jul 18, 2016 7:14 am

Hi Jason,

For these data, p(2) will indeed be estimated as 1, as there is no information about individuals being missed at occasion 2. You may find this biologically unrealistic, but that's what the (limited) data tell you. If you fit a model with constant p, you will also find that the estimated p is 1. As I'm sure you know, the last occasion contains no unique information on p or phi, and only the product phi(2)p(3) can be estimated. Models p(t)p(t), phi(.)p(t) and phi(t)p(.) will have identical deviance and AIC - this is general for 3 occasions.

Generally speaking, although a CJS estimate is technically feasible with 3 occasions, this case is a good example of why it's strongly advised to have more occasions. It's hard to be sure that the apparenty very high p at occassion 2 is real, and even harder to know whether this can be extrapolated to occasion 3. But that's what you can do with such limited data.

Good luck

Morten
Morten Frederiksen
 
Posts: 42
Joined: Thu May 29, 2003 4:53 am
Location: Roskilde, Denmark

Re: 3-year CJS analysis with no 101 encounter history?

Postby Myfanwy » Fri Mar 17, 2017 1:03 am

In a similar situation where individuals are always captured on the first occasion (for example, at an initial release event for a closed marked population), and where no 101 encounter histories occur, will estimated survival always equal relative survival (the number of individuals released divided by the number of individuals captured at the 2nd occasion)?
Myfanwy
 
Posts: 5
Joined: Mon Jan 23, 2017 3:34 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 18 guests