Known fate......nightmare

[Wildlife_Gradstudent]

Hello all,

First post on this forum, thank you all for any time you spend reading this post.

I have been struggling with known fate survival analysis using the MARK program for several months now. I feel as though there are some basic concepts that I just completely missed in the introduction book, but I have no idea how to fix them.

I am attempting to code encounter histories in a monthly interval, over the course of 6 months. as an example, the encounter history is like this:

101010101010

I have 65 individual encounter histories, all coded on individual lines.

Now, this analysis takes into account that there are 6 distinct "study areas" that the individual can belong to, and further, that there are 2 age groups (adult, juvenile), 2 genders (Male and Female) and 2 years of study(2013 and 2014).

I am attempting to code this in such a way that I can have it listed as groups. From all of the group variables listed above, I came up with 48 groups, 8 groups for each study area. groups as
"study area 1, adult, male, 2014",
"study area 1, adult, male, 2013",
"study area 1, adult, female, 2014",
"study area 1, adult, female, 2013",
"study area 1, juvenile, male 2014",
"study area 1, juvenile, male, 2013",
"study area 1, juvenile, female, 2014",
"study area 1, juvenile, female, 2013",

Then repeated these groups for each study area, to bring the total to 48 groups.

Each encounter history is followed by these 48 groups, with a 1 or 0 indicating that the individual belongs to it. So in other words, for each individual they will only have a single "1" designating which group they belong to.

After I did this, I loaded the inp file into MARK and began the tedious task of moving all of these boxes around in the PIM chart.

The simplest model was to just compare survival between years. So, I stacked all of the 2014 groups on the left, and all of the 2013 groups on the right. looks kind of like this:

X
X
X
X
X
X
X

Anyway, after comparing with a consultant who utilized the design matrix to run the analysis, my figures are way off. I asked for their dbf and fpt files to see how they used the design matrix, they replied "try it".....

well, I did try it, but I have no clue what I did wrong or where to even start. This message probably won't be very clear to the reader, so any information you need in order to help me will be provided ASAP. I am just struggling with this so much, and have no idea how to move forward.

[Wildlife_Gradstudent]

formatting did not keep the staggered entry example correct.

it looks more like this

X
...X
X
...X
X
...X
X
...X

[cooch]

You need to read (at least) Chapter 2 (about coding for groups - see for example section 2.1.1), and Chapter 6 (which concerns linear models). In fact, the assumption before you jump into any chapter 8 or higher (which includes the known fate chapters) is that you've mastered everything from chapter 1 -> 7.

If you had, for example, then you'd be able to answer the coding question for yourself. How to setup the .INP file (you don't need 48 columns to specific group, age, sex, gender), or 48 columns in the DM (which you don't seem to be using, which is fairly diagnostic that you might not have assimilated Chapters 1 -> 7.)

[Wildlife_Gradstudent]

Thank you for your response. I have read chapter 2, but I feel like I am confused about how to handle so many different groups. My first attempt at coding these individual encounter histories was formatted as:

encounter history / study area 1 (SA1)/SA2/SA3/SA4/SA5/SA6/Age/Sex/Year

so an example of it was

101010101010 1 0 0 0 0 0 0 1 1 1

This was wrong, so i changed it to

encounter history/ adult male/adult female/juvenile male/juvenile female/Year (1-2014,0-2013), followed by the 6 study area groups.

example:

101010101010 0 0 1 0 1 1 0 0 0 0 0 0

at this point I set up the analysis for 6 encounter intervals, 4 attribute groups, and 1 covariate (year)

now I am just trying to get the analysis to work to assess survival by age, sex, and year

still can't get it to work right. I tried using the PIM chart, just to analyze the groups, and my numbers are way off. I tried using the examples in the book but the issue is that most of the examples regarding the design matrix are for encounter history formats that are not similar to mine, or are for analysis that is not "known fate" .

Is there any kind of resource online that offers a video introduction to this program? or even a powerpoint or keynote presentation?

Thanks for your help anyway, I will just keep reading and hopefully figure something out.

[cooch]

Thank you for your response. I have read chapter 2, but I feel like I am confused about how to handle so many different groups. My first attempt at coding these individual encounter histories was formatted as:

encounter history / study area 1 (SA1)/SA2/SA3/SA4/SA5/SA6/Age/Sex/Year

so an example of it was

101010101010 1 0 0 0 0 0 0 1 1 1

This was wrong, so i changed it to

encounter history/ adult male/adult female/juvenile male/juvenile female/Year (1-2014,0-2013), followed by the 6 study area groups.

example:

101010101010 0 0 1 0 1 1 0 0 0 0 0 0

at this point I set up the analysis for 6 encounter intervals, 4 attribute groups, and 1 covariate (year)

now I am just trying to get the analysis to work to assess survival by age, sex, and year

still can't get it to work right. I tried using the PIM chart, just to analyze the groups, and my numbers are way off. I tried using the examples in the book but the issue is that most of the examples regarding the design matrix are for encounter history formats that are not similar to mine, or are for analysis that is not "known fate" .

Nonsense. The example in the known fate chapter are known fate examples.

You can reduce the dimensions of the problem significantly (in terms of the mechanics) by treating sex and age as individual covariates -- see section 11.7 in chapter 11. In fact, this is *exactly* what is done for the blackduck example in chapter 16 -- 16.4. Look closely at how the .inp file is formatted.

However, all of this is predicated on understanding how things work. There is no substitute for going through -- page by page, example by example - chapters 1 -> 7 (and, for this problem, perhaps followed by chapter 11). In fact, the problem you're describing is very straightforward -- the only complication is having multiple levels of various groups -- which is causing you problems because of the sheer number of parameter structures that you need to manipulate (especially if you use the PIM chart approach). But you can make the problem *much* simpler by getting rid of sex and age as 'factors' treat them as binary covariates, and that will reduce your problem to 4 blue boxes (in the PIM chart vernacular).

Once you have things better understood, you'll be dispensing with the blue boxes anyway, and becoming acquainted with the design matrix (Chapter 6).

Also, when you start the analysis, if you use the individual covariate approach, age and sex are no longer grouping variables, but are specified as individual covariates.

[bvernasco]

Hi,

I recommend you use RMark after figuring out your .inp file. It makes model construction way more streamlined once you figure out the coding. There are lots of good resources on how to learn RMark as well, here is one: http://www.phidot.org/software/mark/rma ... pNotes.pdf

Ben