Blackduck known fate example

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

Blackduck known fate example

Postby jlaake » Mon Feb 25, 2019 8:10 pm

The following was sent to me offlist. Below is my responses to the questions.

I have two quick (I think) RMark questions. I’m using the Known-Fate model to look at survival of rehabilitated black bears throughout a year using a weekly survival interval (i.e., 52 capture histories). I’ve read the RMark workshop pdf, the Known-Fate appendix, the White and Cooch chapters, and followed along with the Blackduck example—which is actually where some the confusion comes in.



In the ‘Blackduck’ example, it states, “factor variables are no longer allowed” and proceeds to change ‘BirdAge’ to a numeric interval variable. I looked through the Phidot message board and did not find any other mention of this. From the data, it appears it should be a group variable, but this example does not specify groups. Is there a reason this example would use an interval age as an individual covariate? Should I have switched from factor variables to numeric?


The change was made so folks wouldn't do something that made no sense. Prior to version 1.6.3 if you used a factor variable without defining it as a group variable, it would use the numeric value of the factor level (1,2...). For this example, the values would be 1 and 2. With 2 levels, this would give the same answer (likelihood, real values etc) as treating it as a 0/1 numeric value but the beta values would be different. With 2 levels of a factor variable there are 2 parameters and they can be either separate parameters (factor) or a slope and intercept. But if you had a factor variable with 3 or more levels, you should get 3 parameters. But before v1.6.3 it would would have just treated it as a numeric variable with values 1,2,3 and you would only estimate 2 parameters - slope and intercept and it would be a silly model.


Also, just to make sure I am on track with interpreting my output summary, I am using several group covariates (sex, region, release location). I have them formatted as factor variables and grouped as I process the data, “bear.processed<-process.data(bdata,model="Known",groups=c("sex","region","location"))”. My output is producing an ‘S’ parameter for each group category (e.g., S.sex: Group:sexF.regionMountains.locationDaniel_Boone 0.9884393; Group:sexM.regionMountains.locationDaniel_Boone 0.9913545) for the respective model.

When you have more than one factor variable to define groups as you have, it will create a group for each unique combination of the factor variables that exists in the data and you will get a real estimate for each group but these may be the same for some groups depending on what model you used. For example, if you used ~1 they would all be the same real estimate.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Return to RMark

Who is online

Users browsing this forum: No registered users and 12 guests