Houston, we have a median c-hat problem here!

[Billy.Requena]

Hello MARK users,

I am dealing with a bitter problem regarding overdispersion.
I have a data set comprising 850 capture-recapture histories, 12 sampling occasions and a multi-model approach, in which individuals are classified as:
(A) females
(B) caring males (males providing parental care)
(C) non-caring males (males not providing parental care)

Transitions between A and any other state are fixed as zero, given individuals cannot change their sex through development. I put males and females data altogether to explicitly construct models comparing apparent survival and recapture rates between sexes (or specifically, between A and C).

Although MARK has some tools to deal with overdispersed data, as the estimation of the median c-hat, which could be used to correct the AICc values and ranking the alternative models, this procedure is recommendable only if the observed factor of overdisperson (c-hat) is no bigger than 3. However, I am facing an observed c-hat of almost 6.5. Furthermore, when I try to simulate the median c-hat by the bootstrap procedure implemented in MARK, it takes too long and freezes in the middle, forcing me to shut down my computer and start over again. In the last attempt, this problem shut the computer down by itself.

Does anyone have a good advice how to proceed now? Or at least a good guess?
Thanks in advance

Gustavo Requena
PhD student - Laboratory of Arthropod Behavior and Evolution
Universidade de São Paulo
São Paulo - SP, Brazil
http://ecologia.ib.usp.br/opilio/gustavo.html

[claudiapenaloza]

Off the top of my head I don't know what to do with the c-hat... but have you thought about analyzing female and male data separately? Are the reproduvtive males and "their" females are always found together, or NEVER found together? If so, this would cause lack of independence in your samples and the overdispersion you are seeing.

Do I have this right MARK people? Is there a better solution?
Maybe give it a try...

claudia

[Billy.Requena]

Hello Claudia (again)

As I mentioned in another post, the species I studied shows exclusive paternal care and females abandon males after oviposition.
I had considered a separate analysis for males and females, but I would like top explicitly compare recapture and survival probabilities of females and males in the non-caring state, comparing models that consider individuals in both categories have the same or different estimates for such parameters. The reason for that is: non-caring males and females walk around searching for food and for mates, while caring males stay at the same spot caring for the offspring. It would be really nice associate behavioral observations to differences among states, don't you think?
But I'm not sure if that is the most correct way to do that...
Thanks in advance

Billy