known fate model and age

[Beatriz Martín]

I'm starting to use Mark with my radiotelemetry data. We have marked animals over nine years (a few individuals every year). I think that mortality in first year (and probably in second one) is bigger than in the rest of years. But I don't know how to incorporate age in my known fate model through the PIM matrix since all animals weren't marked in the same year. I think that it could be imposible and I must take data for each marking year separately but I would prefer to analyses them together because I haven't got many marked individuals.

What can I do?

Thank you
Beatriz

[brant]

I'm starting to use Mark with my radiotelemetry data. We have marked animals over nine years (a few individuals every year). I think that mortality in first year (and probably in second one) is bigger than in the rest of years. But I don't know how to incorporate age in my known fate model through the PIM matrix since all animals weren't marked in the same year. I think that it could be imposible and I must take data for each marking year separately but I would prefer to analyses them together because I haven't got many marked individuals.

Beatriz,

take a look at page 2-7 here for the datafile format (e.g. known fate with an age covariate) and a look here and here (UGA coopunit server) for an additional example of what you are trying to do.

it seems easiest and best (and perhaps the only way possible - i don't know -- i don't use the PIMs) to do what you want using the design matrices. these few sites should provide the technical information you need. as far as the theoretical, some is given in the 2nd article and additional sources can be found (e.g. Kaplan and Meier or Williams et al. text: Analysis and management of animal populations).

cheers,
brant

[Beatriz Martín]

I known the pages you recommend me. Anyway, thank you very much.
I think I have solved the problem in this way: I have considered all animals as marked in the first year of the study and I have codified marking year like a dummy variable to make groups and to look for differences between them. In this way, I can explore age effect and year effect at the same time.

Beatriz.

[quote="brant"][quote="Beatriz Martín"]I'm starting to use Mark with my radiotelemetry data. We have marked animals over nine years (a few individuals every year). I think that mortality in first year (and probably in second one) is bigger than in the rest of years. But I don't know how to incorporate age in my known fate model through the PIM matrix since all animals weren't marked in the same year. I think that it could be imposible and I must take data for each marking year separately but I would prefer to analyses them together because I haven't got many marked individuals.[/quote]

Beatriz,

take a look at page 2-7 [url=http://www.phidot.org/software/mark/docs/book/chap2_2nd_vert.pdf]here[/url] for the datafile format (e.g. known fate with an age covariate) and a look [url=http://www.phidot.org/software/mark/docs/book/chap12_5th_vert.pdf]here[/url] and [url=http://coopunit.forestry.uga.edu/FORS8390/Labs/Week9]here (UGA coopunit server)[/url] for an additional example of what you are trying to do.

it seems easiest and best (and perhaps the only way possible - i don't know -- i don't use the PIMs) to do what you want using the design matrices. these few sites should provide the technical information you need. as far as the theoretical, some is given in the 2nd article and additional sources can be found (e.g. Kaplan and Meier or Williams et al. text: Analysis and management of animal populations).

cheers,
brant[/quote]

[bthompson]

I known the pages you recommend me. Anyway, thank you very much.
I think I have solved the problem in this way: I have considered all animals as marked in the first year of the study and I have codified marking year like a dummy variable to make groups and to look for differences between them. In this way, I can explore age effect and year effect at the same time.

Beatriz.

Hi Beatriz,

If I understand your approach correctly, you are ignoring the staggered entry form of your data by replacing “00s” with “10s” in time intervals preceding the initial marking intervals. If so, this inflates the number of animals at risk in those time intervals, which will inflate their survival estimates (unless no animals died within those intervals).

As an alternative, you might try splitting the encounter histories based on age group, such as 1-year old animals vs. 2-year old and older animals. For example, say you initially marked animal #1 as a 1-year old during year 3 and it died during year 5, and you initially marked animal #2 as a 3-year old during year 4 and it lived to the end of the study (year 9). Here are the typical encounter histories:

/* #1 */ 000010101100000000 1;
/* #2 */ 000000101010101010 1;

Here are the split encounter histories with a group covariate added:

/* #1a */ 000010000000000000 1 0;
/* #1b */ 000000101100000000 1 1;
/* #2 */ 000000101010101010 1 1;

where the group covariate is 0 = 1-year old and 1 = 2-year old and older. Thus, in this example, you would only be splitting those histories where the marked animal was initially marked as a 1-year old. Note that the number of animals at risk for each time interval has not changed. You can do the same for data following the known-fate format. First, here are the data combined into a single group:

known fate group=1;

0 0;
0 0;
1 0;
2 0;
2 1;
1 0;
1 0;
1 0;
1 0;

and then split into 2 groups (age categories), where group 1 = 1-year old and group 2 = 2-year old and older:

known fate group=1;

0 0;
0 0;
1 0;
0 0;
0 0;
0 0;
0 0;
0 0;
0 0;

known fate group=2;

0 0;
0 0;
0 0;
2 0;
2 1;
1 0;
1 0;
1 0;
1 0;

One further comment… If you only marked a few animals each year, you may not have enough marked animals (and hence, animals at risk and animals dying) to detect a difference in survival estimates between ages and across years. That is, you may have quite a few survival estimates of 1.0.

I hope this helps.

Regards,

Bill


P.S. FYI – you should probably be posting questions like this to the “statistics and analysis help” forum.

[Beatriz Martín]

Thank you very much for your suggestions Bill. I think they are very useful for me. The real problem isn't the age at time of marking (because all animals were marked in their first year of life). The problem is that survival is expected to increase with age and changes between anual survival using data without making groups depends on age and not only on year. This is the reason I wanto to separete both effects. Yes, I know that it's possible that I haven't got enough data to detect both effects. But I am specially interested in age effects.

Beatriz

[bthompson]

Thank you very much for your suggestions Bill. I think they are very useful for me. The real problem isn't the age at time of marking (because all animals were marked in their first year of life). The problem is that survival is expected to increase with age and changes between anual survival using data without making groups depends on age and not only on year. This is the reason I wanto to separete both effects. Yes, I know that it's possible that I haven't got enough data to detect both effects. But I am specially interested in age effects.

Beatriz

Hi Beatriz,

If you initially marked all 1-year-olds, then you can still use the approach in my previous example as described for animal #1, but you would apply it to all animals. Thus, you would split encounter histories for each marked animal based on your specified age categories and assign them to the relevant group under the known-fate format (or group covariate if you're using encounter history format).

Take care,

Bill

[Beatriz Martín]

Thank you again.

Beatriz

[bthompson]

Thank you again.

Beatriz

Hi Beatriz,

Actually, if you only marked young animals, I don't think my suggestion will work, either. For example, you will only have young animals marked in year 1 of the study.

Bill

[cooch]

I'm starting to use Mark with my radiotelemetry data. We have marked animals over nine years (a few individuals every year). I think that mortality in first year (and probably in second one) is bigger than in the rest of years. But I don't know how to incorporate age in my known fate model through the PIM matrix since all animals weren't marked in the same year. I think that it could be imposible and I must take data for each marking year separately but I would prefer to analyses them together because I haven't got many marked individuals.

What can I do?

Thank you
Beatriz

Probably no longer of interest, but the basic question(s) you've asked are addressed in the new 'known-fate' chapter - using an approach that is simpler than some of the approaches suggested in replies to your original posting.