Rmark rookie; real vs beta parameters

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

Rmark rookie; real vs beta parameters

Postby j.harv3y » Mon Oct 15, 2018 7:06 am

Hi All,
I seem to have gotten myself very confused over what I'm sure is very straight forward! Could someone please tell me how I can interpret the estimated survival of the population from my CJS model?
I was under the impression that the b values were the most important thing but apparently the 'real' values are back-transformed, so should I be looking to these?

Heres the output;

Output summary for CJS model
Name : Phi(~time)p(~1)

Npar : 6 (unadjusted=4)
-2lnL: 249.2936
AICc : 261.8186 (unadjusted=257.54051)

Beta
estimate se lcl ucl
Phi:(Intercept) 19.2732480 0.0000000 19.2732480 19.273248
Phi:time2013 -14.9518710 0.0000000 -14.9518710 -14.951871
Phi:time2014 -0.3899223 1050.4185000 -2059.2103000 2058.430400
Phi:time2015 -18.1106070 0.0000000 -18.1106070 -18.110607
Phi:time2016 -17.4604770 0.0000000 -17.4604770 -17.460477
p:(Intercept) 1.2694602 0.2162026 0.8457031 1.693217


Real Parameter Phi

2012 2013 2014 2015 2016
2012 1 0.9868925 1 0.7618123 0.8596964
2013 0.9868925 1 0.7618123 0.8596964
2014 1 0.7618123 0.8596964
2015 0.7618123 0.8596964
2016 0.8596964


Real Parameter p

2013 2014 2015 2016 2017
2012 0.7806503 0.7806503 0.7806503 0.7806503 0.7806503
2013 0.7806503 0.7806503 0.7806503 0.7806503
2014 0.7806503 0.7806503 0.7806503
2015 0.7806503 0.7806503
2016 0.7806503

Thanks so much in advance!
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am

Re: Rmark rookie; real vs beta parameters

Postby cooch » Mon Oct 15, 2018 8:40 am

j.harv3y wrote:Hi All,
I seem to have gotten myself very confused over what I'm sure is very straight forward! Could someone please tell me how I can interpret the estimated survival of the population from my CJS model?
I was under the impression that the b values were the most important thing but apparently the 'real' values are back-transformed, so should I be looking to these?

Heres the output;

Output summary for CJS model
Name : Phi(~time)p(~1)

Npar : 6 (unadjusted=4)
-2lnL: 249.2936
AICc : 261.8186 (unadjusted=257.54051)

Beta
estimate se lcl ucl
Phi:(Intercept) 19.2732480 0.0000000 19.2732480 19.273248
Phi:time2013 -14.9518710 0.0000000 -14.9518710 -14.951871
Phi:time2014 -0.3899223 1050.4185000 -2059.2103000 2058.430400
Phi:time2015 -18.1106070 0.0000000 -18.1106070 -18.110607
Phi:time2016 -17.4604770 0.0000000 -17.4604770 -17.460477
p:(Intercept) 1.2694602 0.2162026 0.8457031 1.693217


Real Parameter Phi

2012 2013 2014 2015 2016
2012 1 0.9868925 1 0.7618123 0.8596964
2013 0.9868925 1 0.7618123 0.8596964
2014 1 0.7618123 0.8596964
2015 0.7618123 0.8596964
2016 0.8596964


Real Parameter p

2013 2014 2015 2016 2017
2012 0.7806503 0.7806503 0.7806503 0.7806503 0.7806503
2013 0.7806503 0.7806503 0.7806503 0.7806503
2014 0.7806503 0.7806503 0.7806503
2015 0.7806503 0.7806503
2016 0.7806503

Thanks so much in advance!


Read chapter 6 of the MARK book - carefully.

http://www.phidot.org/software/mark/doc ... /chap6.pdf
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: Rmark rookie; real vs beta parameters

Postby jlaake » Mon Oct 15, 2018 11:48 am

Something isn't looking right with your estimates. Is this the model with minimum AIC? I'd be surprised if it is. You have at least 2 parameters at boundaries. One is the intercept, presumably survival in 2012. Try using ~-1 + time for the Phi formula.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Rmark rookie; real vs beta parameters

Postby j.harv3y » Mon Oct 15, 2018 4:43 pm

Thanks so much both for the help!

jlaake wrote:Something isn't looking right with your estimates. Is this the model with minimum AIC? I'd be surprised if it is. You have at least 2 parameters at boundaries. One is the intercept, presumably survival in 2012. Try using ~-1 + time for the Phi formula.


Thanks jlaake, thats incredibly helpful to know... Having problems with understanding what exactly the problem is with this model. How did you know it is two parameters at boundaries? Additionally could you explain the use of the -1 + time please? :)

Thanks so much again,

Jess
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am

Re: Rmark rookie; real vs beta parameters

Postby jlaake » Mon Oct 15, 2018 8:21 pm

I could tell because the output showed 6 parameters but the unadjusted count from MARK is 4. The difference is 2 wich is the number at a boundary in this case. The -1+time says to remove the intercept and then each time beta is estimated separately.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Rmark rookie; real vs beta parameters

Postby j.harv3y » Tue Oct 16, 2018 2:26 pm

jlaake wrote:I could tell because the output showed 6 parameters but the unadjusted count from MARK is 4. The difference is 2 wich is the number at a boundary in this case. The -1+time says to remove the intercept and then each time beta is estimated separately.


Jlaake, you're a superhero, thanks so much for explaining, I really appreciate it!

So I tried what you suggested and the problem is still there with the parameters- annoyingly there's an additional parameter this time!

Could you perhaps suggest a reason for this, is there anyway of explicitly finding the parameters which are problematic?

STOP NORMAL EXIT

Note: only 7 parameters counted of 10 specified parameters

AICc and parameter count have been adjusted upward


Output summary for CJS model
Name : Phi(~-1 + time)p(~time)

Npar : 10 (unadjusted=7)
-2lnL: 246.5837
AICc : 267.994 (unadjusted=261.28815)

Beta
estimate se lcl ucl
Phi:time2012 25.2404510 1.917850e+04 -3.756462e+04 3.761510e+04
Phi:time2013 4.8468071 3.708287e+00 -2.421435e+00 1.211505e+01
Phi:time2014 20.7564340 7.737118e+03 -1.514400e+04 1.518551e+04
Phi:time2015 0.9977356 3.593416e-01 2.934259e-01 1.702045e+00
Phi:time2016 0.8220227 1.548840e+00 -2.213703e+00 3.857748e+00
p:(Intercept) 1.2809364 5.055258e-01 2.901059e-01 2.271767e+00
p:time2014 -0.3941745 6.282424e-01 -1.625530e+00 8.371805e-01
p:time2015 -0.0413504 6.190753e-01 -1.254738e+00 1.172037e+00
p:time2016 0.7215290 7.952128e-01 -8.370882e-01 2.280146e+00
p:time2017 6.7085375 1.356843e+03 -2.652704e+03 2.666121e+03


Real Parameter Phi

2012 2013 2014 2015 2016
2012 1 0.9922078 1 0.7306131 0.6946655
2013 0.9922078 1 0.7306131 0.6946655
2014 1 0.7306131 0.6946655
2015 0.7306131 0.6946655
2016 0.6946655


Real Parameter p

2013 2014 2015 2016 2017
2012 0.7826091 0.7082215 0.7754919 0.8810557 0.9996611
2013 0.7082215 0.7754919 0.8810557 0.9996611
2014 0.7754919 0.8810557 0.9996611
2015 0.8810557 0.9996611
2016 0.9996611

Aside from the parameters of phi you suggested I had run 6 different models. All the models in which run without this error have phi as constant (~1), however they have the highest AICC values.

model npar AICc DeltaAICc weight Deviance
3 Phi(~time)p(~1) 6 261.8186 0.000000 0.57931377 82.88228
5 Phi(~time)p(~Time) 7 262.8864 1.067753 0.33966777 81.77062
7 Phi(~-1 + time)p(~time) 10 267.9940 6.175396 0.02642065 80.17242
4 Phi(~time)p(~time) 10 267.9940 6.175406 0.02642052 80.17242
6 Phi(~1)p(~Time) 3 268.8048 6.986209 0.01761480 96.24625
1 Phi(~1)p(~1) 2 270.3329 8.514311 0.00820455 99.84841
2 Phi(~1)p(~time) 6 272.8267 11.008120 0.00235793 93.89039


What would your thoughts be on the next step?

Again, your advice is invaluable, cant thank you enough!!
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am

Re: Rmark rookie; real vs beta parameters

Postby jlaake » Tue Oct 16, 2018 3:18 pm

You do need to take Evan's advice and read his electronic book. If you did you would know that only the product phi (t)p (t) is estimable for last occasion with time variation in phi and p. If you compare models phi (-1+time )p (.) and phi (time)p (.) you will see that they will have same aic etc but the former will have 2 parameters at a boundary with beta being very large positive or very large negative value and a std error that is either very large or 0.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Rmark rookie; real vs beta parameters

Postby cooch » Wed Oct 17, 2018 8:42 am

If you're going to start reading 'the book', then given your question(s), I'd suggest chapter 1, 3-6, before you get anywhere near your own data. Those 5 chapters are 'minimum sufficient' for you to have a fair idea what you're doing.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: Rmark rookie; real vs beta parameters

Postby j.harv3y » Fri Oct 19, 2018 8:07 am

Hi all,
Thanks so much for the advice, have now read the chapters suggested!
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am


Return to RMark

Who is online

Users browsing this forum: No registered users and 8 guests