Page **1** of **1**

### Sparse data and no recapture

Posted:

**Mon Jan 24, 2011 11:40 pm**
by **fionafoo**

Hi everybody,

i am carrying out a survey to estimate the population density of rattus argentiventer by using the software DENSITY 4.4

This live trapping was done for two season.

The 1st season, i did not have any recapture while the 2nd season i do have recapture which is always in the same site (cage) and for both season my data is sparse.

i understand that since my recapture is always in the same cage there is no information about the scale of movement and detection.

So i would like to know if there is any possible way for me to still being able to calculate the population density using this DENSITY with my data.

### Re: Sparse data and no recapture

Posted:

**Tue Jan 25, 2011 12:09 am**
by **murray.efford**

fionafoo wrote:Hi everybody,

i am carrying out a survey to estimate the population density of rattus argentiventer by using the software DENSITY 4.4

This live trapping was done for two season.

The 1st season, i did not have any recapture while the 2nd season i do have recapture which is always in the same site (cage) and for both season my data is sparse.

i understand that since my recapture is always in the same cage there is no information about the scale of movement and detection.

So i would like to know if there is any possible way for me to still being able to calculate the population density using this DENSITY with my data.

Fiona

With sparse data like this you can only estimate density by making additional assumptions, but it _is_ possible using ML SECR. If you are willing to assume that movements were similar between seasons then you can fit a pooled, constant model for detection, while allowing density to vary. To do this in DENSITY set up the seasons as different 'sessions' in the one dataset and on the Options | ML SECR page tick 'Use between session model' on the 'Between session model' tab; the default between-session model is exactly what you want (session-specific density, constant g0 and sigma). [You can achieve the same in 'secr' (function secr.fit) with model = list(D~session, g0~1, sigma~1)]. You might also try fitting a model with session-specific g0 and constant sigma.

I guess that you used single-catch traps, and may feel uncomfortable about using ML SECR models for multi-catch traps. This really isn't a worry for density estimation, and from the sparseness of your data I guess the traps weren't saturated (See Efford Borchers & Byrom 2009).

Hope this works

Murray

### Re: Sparse data and no recapture

Posted:

**Tue Jan 25, 2011 12:12 am**
by **murray.efford**

Oops - I answered a different question, sorry. I skipped over the 'not' in your first season recaptures. I'm sorry, but with no within-season movements you're lost.

Murray

### Re: Sparse data and no recapture

Posted:

**Wed Jan 26, 2011 8:49 am**
by **fionafoo**

Thank you very much.

### Re: Sparse data and no recapture

Posted:

**Wed Jan 26, 2011 11:10 am**
by **fionafoo**

I am sorry, i don't quite understand the meaning of no within season movement here, because in the software it did not ask me to key in any movements.

### Re: Sparse data and no recapture

Posted:

**Wed Jan 26, 2011 2:23 pm**
by **murray.efford**

For each occasion on which an animal was captured, the data include the trap location. For recaptures this may be the same as the previous capture (no movement) or different from the previous capture (implying movement). That is all I meant; spatially explicit capture-recapture models do not deal with movement directly. They use a detection function with a spatial scale parameter (sigma) to represent the result of movement: individuals tend to be detected over an area but more often near their-home range centres. If the data include no recaptures in different traps, whether by chance or because traps were too far apart, then there is no information with which to estimate sigma.

To complicate matters slightly: if you actually 'know' sigma (e.g. by assuming it is the same as in another study) then you may be able to fit the model with it fixed at this value (I can't remember whether this works or the program aborts immediately when it finds no movements). In Density you would fix sigma in Options | ML SECR by setting the Initial values to 'Manual', double-clicking on the link function for Sigma until it is 'Fixed', and entering the value in the Initial column. The estimate of density will be very sensitive to the particular value you choose for sigma, and I don't recommend this route, but it may help you understand what is going on.

Sadly, sparse data will remain uninformative.

Murray

### Re: Sparse data and no recapture

Posted:

**Fri Jan 28, 2011 3:46 am**
by **fionafoo**

Thanks for the information

### Re: Sparse data and no recapture

Posted:

**Fri Jan 28, 2011 12:14 pm**
by **fionafoo**

hi

I tried the advise you gave me by adding assuming the sigma.

I obtained the sigma value by running the season 2 data using the ML model. After achieving the sigma value, i assume the sigma value is the same and run a combination my data for season 1 and season 2 to estimate my density. This is just a trial run as you did mention that this method is not recommended in your previous reply .

I would like to know if i may assume my sigma this way, and since i am using single-catch trap will it affect my estimated density by using the ML model.

### Re: Sparse data and no recapture

Posted:

**Fri Jan 28, 2011 2:43 pm**
by **murray.efford**

As I understand it, your only recaptures are in the same trap. You will not be able to estimate sigma or density from these data.

Fixing sigma (as in my last post) is not the same as pooling data across sessions to estimate sigma (as suggested in my first post). I mentioned the technical possibility of fixing a parameter (sigma) only for completeness, but maybe this was confusing.

Murray