Data of A. E. Byrom from a study of stoats (Mustela erminea) in New Zealand. Individuals were identified from DNA in hair samples.
stoatCH stoat.model.HN stoat.model.EX
The data are from a pilot study of stoats in red beech (Nothofagus fusca) forest in the Matakitaki Valley, South Island, New Zealand. Sticky hair-sampling tubes (n = 94) were placed on a 3-km x 3-km grid with 500-m spacing between lines and 250-m spacing along lines. Tubes were baited with rabbit meat and checked daily for 7 days, starting on 15 December 2001. Stoat hair samples were identified to individual using DNA microsatellites amplified by PCR from follicular tissue (Gleeson et al. 2010). Six loci were amplified and the mean number of alleles was 7.3 per locus. Not all loci could be amplified in 27% of samples. A total of 40 hair samples were collected (Gleeson et al. 2010), but only 30 appear in this dataset; the rest presumably did not yield sufficient DNA for genotyping.
The data are provided as a single-session
‘stoatCH’. Hair tubes are ‘proximity’ detectors which allow
an individual to be detected at multiple detectors on one occasion
(day), but there are no multiple detections in this dataset and for
historical reasons the data are provided as detector type ‘multi’. Two
pre-fitted models are included:
|stoat.model.HN||fitted secr model -- null, halfnormal detection function|
|stoat.model.EX||fitted secr model -- null, exponential detection function|
The log-likelihood values reported for these data by
differ by a constant from those published by Efford et al. (2009)
because the earlier version of DENSITY used in that analysis did not
include the multinomial coefficient, which in this case is log(20!) or
about +42.336. The previous analysis also used a coarser habitat mask
than the default in secr (32 x 32 rather than 64 x 64) and this
slightly alters the log-likelihood and
Fitting the hazard-rate detection function previously required the shape parameter z (or b) to be fixed, but the model can be fitted in secr without fixing z. However, the hazard rate function can cause problems owing to its long tail, and it is not recommended. The check on the buffer width, usually applied automatically on completion of secr.fit, causes an error and must be suppressed with biasLimit = NA (see Examples).
Gleeson et al. (2010) address the question of whether there is enough variability at the sampled microsatellite loci to distinguish individuals. The reference to 98 sampling sites in that paper is a minor error (A. E. Byrom pers. comm.).
Efford, M. G., Borchers D. L. and Byrom, A. E. (2009) Density estimation by spatially explicit capture-recapture: likelihood-based methods. In: D. L. Thomson, E. G. Cooch and M. J. Conroy (eds) Modeling Demographic Processes in Marked Populations. Springer, New York. Pp. 255–269.
Gleeson, D. M., Byrom, A. E. and Howitt, R. L. J. (2010) Non-invasive methods for genotyping of stoats (Mustela erminea) in New Zealand: potential for field applications. New Zealand Journal of Ecology 34, 356–359. Available on-line at https://newzealandecology.org/nzje/2936/.
summary(stoatCH) ## Not run: stoat.model.HN <- secr.fit(stoatCH, buffer = 1000, detectfn = 0) # this generates an error unless we use biasLimit = NA # to suppress the default bias check stoat.model.EX <- secr.fit(stoatCH, buffer = 1000, detectfn = 2) confint(stoat.model.HN, "D") ## Profile likelihood interval(s)... ## lcl ucl ## D 0.01275125 0.04055662 ## End(Not run) ## plot fitted detection functions xv <- seq(0,800,10) plot(stoat.model.EX, xval = xv, ylim = c(0,0.12), limits = FALSE, lty = 2) plot(stoat.model.HN, xval = xv, limits = FALSE, lty = 1, add = TRUE) ## review density estimates collate(stoat.model.HN, stoat.model.EX, realnames = "D", perm = c(2,3,4,1)) modelAverage(stoat.model.HN, stoat.model.EX, realnames = "D")
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.