Why are the survival rates >1 when using mrOpen in the package FSA?

[qoswald]

I am currently doing some population analyses with the package "FSA" in R. By using the mrOpencommand, I want to get the survival rate. My rawdata is a simple table with one row per indidivual, one column per sample date and values of 0 and 1 (for not capured or captured during that respective sampling).

id total.captures date1 date2 date3 etc 1 3 1 1 1 ... 2 1 1 0 0 ...

This is the exact code:

hold.data<-capHistSum(data, cols2use = c(3:13)) est.data<-mrOpen(hold.data) summary(est.data) confint(est.data)

It seems to work out, as I get the tables and summaries with all the parameters. See this example:

However, there's a problem with the survival estimate phi. The phi value is not between 0 and 1, but in some cases, exceeds 1.

Any idea, what went wrong here?

Thanks, Pia

[jcdoll79]

Hi Pia,

When phi > 1, it typically indicates an over-parameterized model (i.e,, the data don't support the underlying model that is being estimated). Would you be willing to share a subset of your data that recreates the issue? You can also email me the data directly, if you would like, jason.doll@fmarion.edu.

Jason

[droglenc]

Pia,

I agree with Jason. Several of the parameters are poorly estimated here (several infinities for SE, SE that are very large). I suspect few recaptured fish, overall or per recapturing event.

[qoswald]

Thanks for your fast answers! @jcdoll79 I'll sent you a mail.

Pia

[jcdoll79]

Hi Pia, Thanks for sending me your data to investigate the issue. @droglenc suspicions are correct, there are too few recaptured fish to get good estimates of survival or abundance. The only suggestion I have is to try a Bayesian approach towards fitting the open Jolly-Seber model where you can specify prior information to provide more realistic estimates (this would require coding the model yourself). You might also try an individual Cormack-Jolly-Seber model. The CJS model will estimate survival but it doesn't directly estimate population size. As of now, FSA does not support CJS models. Jason

[qoswald]

Hi Jason,

thank you so much, again, for helping me with this issue.

It seems I need to dive deeper into Bayesian approaches.

Just to make sure, did I understand you correctly that with my data also the estimated population size is not valid?

Or was this for the survival estimates only?

Have a nice Easter!

Best wishes,

Pia

Von: Jason Doll @.> Gesendet: Donnerstag, 14. April 2022 21:25 An: fishR-Core-Team/FSA @.> Cc: Pia Oswald @.>; Author @.> Betreff: Re: [fishR-Core-Team/FSA] Why are the survival rates >1 when using mrOpen in the package FSA? (Issue #92)

Hi Pia, Thanks for sending me your data to investigate the issue. @droglenc https://github.com/droglenc suspicions are correct, there are too few recaptured fish to get good estimates of survival or abundance. The only suggestion I have is to try a Bayesian approach towards fitting the open Jolly-Seber model where you can specify prior information to provide more realistic estimates (this would require coding the model yourself). You might also try an individual Cormack-Jolly-Seber model. The CJS model will estimate survival but it doesn't directly estimate population size. As of now, FSA does not support CJS models. Jason

— Reply to this email directly, view it on GitHub https://github.com/fishR-Core-Team/FSA/issues/92#issuecomment-1099550671 , or unsubscribe https://github.com/notifications/unsubscribe-auth/ANXSOCGMYSA5BLAAQWRKZJDVFBWJDANCNFSM5TNRKJYQ . You are receiving this because you authored the thread.Message ID: @.***>

[jcdoll79]

Hi Pia, That is correct. The population estimates and the survival estimates are not valid and should not be interpreted. Jason

[qoswald]

Hi Jason,

okay, I will try another approach.

Thanks for your help!

Best wishes,

Pia

Von: Jason Doll @.> Gesendet: Samstag, 16. April 2022 20:43 An: fishR-Core-Team/FSA @.> Cc: Pia Oswald @.>; Author @.> Betreff: Re: [fishR-Core-Team/FSA] Why are the survival rates >1 when using mrOpen in the package FSA? (Issue #92)

Hi Pia, That is correct. The population estimates and the survival estimates are not valid and should not be interpreted. Jason

— Reply to this email directly, view it on GitHub https://github.com/fishR-Core-Team/FSA/issues/92#issuecomment-1100729711 , or unsubscribe https://github.com/notifications/unsubscribe-auth/ANXSOCCPORZEXNAKL56P3RDVFMC2FANCNFSM5TNRKJYQ . You are receiving this because you authored the thread.Message ID: @.***>