Some variance-covariance matrix elements were NA

[Dominik Marchowski]

Hello,

I would like to ask for possible solutions to my problem. 

After using the following formula: "ds (data = data, key =" hn ", formula = ~ log (size) + Obs, cutpoints = bins, convert.units = conversion.factor) " I got the following message:

" Some variance-covariance matrix elements were NA, possible numerical problems; only estimating detection function. "

AIC was lower than in the formula" ~ log (size) "only, so I should use this formula, but unfortunately it is not possible due to the problem described above. I can say that in similar databases, such a message has never appeared before with the formula " log (size)+Obs.

Does anyone have any idea what this may be the result of? I would just like to say that the formula "~ log (size)" and using the variable sea.state ("~ log (size) + sea.state") works fine. There is also no NA among the observer codes in Obs variable.

All the best,

Dominik

[Eric Rexstad]

Greetings Dom, nice to hear from you again.

To diagnose this problem, you will need to look closely at your data.  There is nothing inherently wrong with the model specification.  However for some data sets support for fitting such a model might be problematic.  Remember distance is also tacitly part of your detectability model.

I suggest you do some descriptive analysis of the data set that is causing this problem.  Specifically, what is the distribution of detection distances for different group sizes detected for each of the observers?  I suspect that for one (or more observers) there is a poor distribution of detected group sizes.  I would think these descriptive plots will point to the difficulty reported by the ds()​ function.

You may then find that you will use the model that includes only group size even if it has an inferior AIC score.  My suspicion is that because of pooling robustness, the density estimates of your birds may not be wildly different for the competing models.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Dominik Marchowski <march...@gmail.com>
Sent: 25 January 2022 11:45
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Some variance-covariance matrix elements were NA

 

--
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/CAPbpb8S_AbCkPg5Pw05iPxvZRj6hqBkMFhaupi9_c4SgkjOc4g%40mail.gmail.com.

[Dominik Marchowski]

---------- Forwarded message ---------
Od: Dominik Marchowski <march...@gmail.com>
Date: wt., 25 sty 2022 o 18:05
Subject: Re: [distance-sampling] Some variance-covariance matrix elements were NA
To: Eric Rexstad <Eric.R...@st-andrews.ac.uk>

Thank you Eric for your explanation, I will therefore continue to export my database.

Best,

Dominik

--

Dominik Marchowski, Ph.D.

Museum and Institute of Zoology, Polish Academy of Sciences. 

Wilcza 64, 00-679 Warszawa, Poland. +48.668.139.139

https://www.researchgate.net/profile/Dominik_Marchowski

--

Dominik Marchowski, Ph.D.

Museum and Institute of Zoology, Polish Academy of Sciences. 

Wilcza 64, 00-679 Warszawa, Poland. +48.668.139.139

https://www.researchgate.net/profile/Dominik_Marchowski