Some variance-covariance matrix elements were NA

26 views
Skip to first unread message

antom...@gmail.com

unread,
Mar 9, 2022, 8:33:39 AM3/9/22
to distance-sampling
Dear list, 
Hi! Some time ago I 've  worked with a data set using Rdistance (see attached) and I've got this models:
 summarize_ds_models(ad10.hn,ad10.hn.s,ad10.hn.sr,ad10.hr,ad10.hr.s,ad10.hr.sr)
                 Model Key function              Formula C-vM p-value $\\hat{P_a}$
2  \\texttt{ad10.hn.s}  Half-normal                ~size   0.28617840    0.6809283
3 \\texttt{ad10.hn.sr}  Half-normal ~size + Region.Label   0.34085468    0.6751861
4    \\texttt{ad10.hr}  Hazard-rate                   ~1   0.94819319    0.7176434
5  \\texttt{ad10.hr.s}  Hazard-rate                ~size   0.96843811    0.7207349
6 \\texttt{ad10.hr.sr}  Hazard-rate ~size + Region.Label   0.91570747    0.7166309
1    \\texttt{ad10.hn}  Half-normal                   ~1   0.06288654    0.5934467
  se($\\hat{P_a}$) $\\Delta$AIC
2       0.04449997     0.000000
3       0.04406206     2.003515
4       0.02877457    36.015255
5       0.02861308    37.335866
6       0.02942843    38.311241
1       0.04211494    47.559646

According to AIC I selected the ad10.hn.s model. But now, I am running these models again and I am getting this warning with the selected model: 

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

Are there any changes in the package that might be influencing these results? 
Thanks!
Antonela
4km adultos_2010_febrero.txt

Eric Rexstad

unread,
Mar 9, 2022, 9:50:27 AM3/9/22
to distance...@googlegroups.com
From: Eric Rexstad <Eric.R...@st-andrews.ac.uk>
Sent: 09 March 2022 13:57
To: antom...@gmail.com <antom...@gmail.com>; distance-sampling <distance...@googlegroups.com>
Subject: Re: {Suspected Spam} [distance-sampling] Some variance-covariance matrix elements were NA
 
Antonela

I'm sure you meant to say you were using the Distance package, not the Rdistance package.

Do you remember what version of the Distance package you were using previously and the version of the package that produces the warning message?  Are you working with versions available on CRAN or are you working with the package you downloaded from Github?  There was a version of the Distance (and mrds) package uploaded to CRAN last August, you can read about some of the changes in this post:

mrds 2.2.5 and Distance 1.0.3 Posted 08 July 2021 by Dave Miller. R packages mrds 2.2.5 and Distance 1.0.3 are now on CRAN.. New versions of mrds and Distance are now available on CRAN. There are a couple of big new features and improvements so I’ve added a bit more description here about them.
and this post
Further optimisation improvement details Posted 01 August 2021 by Dave Miller. The newest version of our R package mrds (2.2.5), used for fitting detection functions by Distance, has some improvements to its optimisation system.Here are some more details of recent changes and what you might need to watch out for.

Which of the models in the table of candidate models is producing this warning message?

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of antom...@gmail.com <antom...@gmail.com>
Sent: 09 March 2022 13:33
To: distance-sampling <distance...@googlegroups.com>
Subject: {Suspected Spam} [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/f19f093a-7830-4933-992c-bc52f4ea81b0n%40googlegroups.com.

Antonela Marozzi

unread,
Mar 9, 2022, 1:48:14 PM3/9/22
to Eric Rexstad, distance-sampling
Hi Eric, I am working with versions downloaded from the CRAN. The old version was Distance 1.0.0 and the version that is now on my computer is  Distance 1.0.3. I have checked the models in an old computer with the old version of the Distance package and the models run ok. It seems to be some change between those versions. 
Regards
Antonela

El mié, 9 mar 2022 a las 10:57, Eric Rexstad (<Eric.R...@st-andrews.ac.uk>) escribió:
Antonela

I'm sure you meant to say you were using the Distance package, not the Rdistance package.

Do you remember what version of the Distance package you were using previously and the version of the package that produces the warning message?  Are you working with versions available on CRAN or are you working with the package you downloaded from Github?  There was a version of the Distance (and mrds) package uploaded to CRAN last August, you can read about some of the changes in this post:

mrds 2.2.5 and Distance 1.0.3 Posted 08 July 2021 by Dave Miller. R packages mrds 2.2.5 and Distance 1.0.3 are now on CRAN.. New versions of mrds and Distance are now available on CRAN. There are a couple of big new features and improvements so I’ve added a bit more description here about them.
and this post
Further optimisation improvement details Posted 01 August 2021 by Dave Miller. The newest version of our R package mrds (2.2.5), used for fitting detection functions by Distance, has some improvements to its optimisation system.Here are some more details of recent changes and what you might need to watch out for.

Which of the models in the table of candidate models is producing this warning message?

Subject: {Suspected Spam} [distance-sampling] Some variance-covariance matrix elements were NA
 
--

Eric Rexstad

unread,
Mar 9, 2022, 1:52:52 PM3/9/22
to Antonela Marozzi, distance-sampling
Yes Antonela.  There were changes to the optimisation routines in Distance 1.0.3 as described in the two blog posts I linked in my previous emails.
From: Antonela Marozzi <antom...@gmail.com>
Sent: 09 March 2022 18:48
To: Eric Rexstad <Eric.R...@st-andrews.ac.uk>
Cc: distance-sampling <distance...@googlegroups.com>
Subject: Re: {Disarmed} Re: {Suspected Spam} [distance-sampling] Some variance-covariance matrix elements were NA
 
Reply all
Reply to author
Forward
0 new messages