abundance.est in dsm model

[olga30dec]

Dear all, 

I am struggling to understand how to access response variable abundance.est in dsm model. At the moment I am fitting  family = quasipoisson(link = "log")

However model diagnostics shows that this distribution family does not fit well.

I tried to fit gamma but I get error: Non-Positive values not allowed for the gamma. 

Where do I get  non positive values from? I am trying to inspect my response variable, but I cannot find the way to access it.

Many thanks for your help,

Olga

[Eric Rexstad]

Olga

The short answer to your question regarding the error arising from trying to fit a gamma is that 0 is a non-positive value.  Therefore segments without detections will have 0 as the abundance; the gamma does not like this.

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/63966509-7e2e-4f28-bf58-cf3d7ac1ede3n%40googlegroups.com.

-- 
Eric Rexstad
Centre for Ecological and Environmental Modelling
University of St Andrews
St Andrews is a charity registered in Scotland SC013532

[Olga Lyashevska]

Thank you very much Eric for this clarification. This makes sense. What distribution within dsm (or mgcv) can deal with zero-inflation? Segmentation introduces zeros, the number of which which depend on the size of segments. The only option I see is zero-inflated Poisson, however adundance.est is not count. What would be the best way to address this?  Thanks for your help! 

[Eric Rexstad]

Olga

You might consider the Tweedie distribution, as shown in

http://examples.distancesampling.org/dsm-line-dolphins/mexico-analysis.html

To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/41d890cf-c6d5-4c9b-b840-b507899f5e23n%40googlegroups.com.