Stratifying Data by year for annual abundance estimates.

[Fay Frost]

Hi all, 

I am very new to distance and I am looking to get some density/abundance estimates of some distance data from 2006-2021. The data is almost yearly, though in some years there is more than one survey done.  

I also have lots of covariate information such as temperature, habitat etc which at some point I would like to include in the model selection process. 

I am wondering what the best approach is for setting up an analysis where I want to get annual estimates? Should I stratify by year or use a post-stratify, I am not sure I know what the difference is? And does this choice change if I want to include covariates?

Any advice will be welcomed. 

Best wishes, 

Fay

[Eric Rexstad]

Fay, welcome to the distance sampling group.

I presume you are using the Distance for Windows software rather than the Distance R package; not that this changes the answer to your questions, but just for clarity.

There isn't a universal answer to your question, but as a starting point, seems most straightforward to treat your years as strata; setting aside the question of what to do for years with multiple surveys.  I'm also ignoring the possibility that your surveys have geographical strata as well as multiple years.

An important feature if you wish to produce year-specific estimates is that the detection function not be pooled across years (strata).  If there are an adequate number of detections in each year, then simply ask for detection function (as well as encounter rate) to be estimated separately for each year.  It is the situation where there are inadequate numbers of detections in some years that decisions regarding how to proceed become more complex.

Regarding inclusion of covariates, recognise that conventional distance sampling only incorporates covariates in the detection function.  Don't know the species of interest, but you might ask whether covariates influence the shape of the detection function; exploratory data analysis required before diving into the pool of detection function modelling with covariates.  Recognise also the "pooling robustness" property of distance sampling whereby unbiased estimates of density will still be produced even if some sources of variation in detectability are not included in the detection function modelling.  For sound distance sampling data, the estimates of density are not appreciably affected by inclusion of covariates in the detection function.

That's a three paragraph consulting session.  See how far that takes you and write back when the situation warrants.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Fay Frost <fayfr...@gmail.com>
Sent: 17 March 2022 15:12
To: distance-sampling <distance...@googlegroups.com>
Subject: {Suspected Spam} [distance-sampling] Stratifying Data by year for annual abundance estimates.

 

--
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/8566cc4c-5b51-4af1-b516-58752f7c5177n%40googlegroups.com.

[Fay Frost]

Hi Eric, 

Thank you so much for a speedy reply. It was very helpful.

I am currently doing work in Distance for Windows but I would like to move onto R at some point, especially given that I do have geographical strata that I am currently ignoring. 

 I have a couple of follow up questions though, if you have time. 

1.) How many detections are sufficient per strata, I have between 40-70, do you think this is adequate?

2.) I have attached an image below for my model set up,  have I estimated detection and density how you suggested? I.e. separately for each strata.

3.) I have done the analysis using both stratify by year and also post-stratify by year, I seem to get the same abundance/density estimates. What is the difference between the two?

Lastly, a lot of my analyses give an error message

**Warning: Parameters are being constrained to obtain monotonicity**

I have watched the distance course, and I am still unsure what this means and how to fix it. Could you please help?

Thanks again for all your help!

Fay 

[Eric Rexstad]

Fay

Your questions:

• that number of detections per year is likely to be adequate.  Look for indications of really poor precision in the detection function parameters as a signal of possibly inadequacy
• yes, the check boxes show that separate detection functions will be fitted for each stratum.  The "global" density estimate is unlikely to be informative; it is an average of density across the 15 years; likely you can uncheck that
• if the criterion you have asked for your post-stratification is identified at the stratum level in your data, the results will be identical.  Post-stratification is used in fairly unusual situations where the criterion for stratification was not known at the time the survey was designed.  An example would be post-stratification by males/females; a design couldn't be created knowing "strata" where males or females would be found.

Many people ask about monotonicity.  Note it is a warning​​ not an error​​.  Software is "conflicted" between trying to produce the best fitting model (which may have detectability increasing as a function of distance over some range of distances) and the rule that states "detectability is a non-increasing function of perpendicular distance from the transect".  For some of your analyses, the software cannot achieve both of these things and it is telling you that it has favoured the second (non-increasing function) rule over the first (best fit) rule.  No real "fix" is required on your part.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Fay Frost <fayfr...@gmail.com>

Sent: 18 March 2022 11:00
To: distance-sampling <distance...@googlegroups.com>
Subject: {Suspected Spam} Re: {Suspected Spam} [distance-sampling] Stratifying Data by year for annual abundance estimates.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/c19c79f1-5f5d-413f-b4a5-9cc6fc17d525n%40googlegroups.com.