Estimates the Adapted Pair Correlation Function (PCF) of a pattern together
with a pointwise critical envelope based on distances and ratios calculated
by pat2dists().
Arguments
- dists
An object of class dists. Usually created by
pat2dists()- r
A step size or a vector of values for the argument r at which g(r) should be evaluated.
- r_max
maximum value for the argument r.
- kernel
String. Choice of smoothing kernel (only the "epanechnikov" kernel is currently implemented).
- stoyan
Bandwidth coefficient (smoothing the Epanechnikov kernel). Penttinen et al. (1992) and Stoyan and Stoyan (1994) suggest values between 0.1 and 0.2.
- n_rank
Rank of the value amongst the n_sim simulated values used to construct the envelope. A rank of 1 means that the minimum and maximum simulated values will be used. Must be >= 1 and < n_sim/2. Determines together with
n_siminpat2dists()the alpha level of the envelope. Ifalphaandn_simare fix, n_rank can be calculated by(n_sim+1)*alpha/2eg.(199+1) * 0.05/2 = 5
Value
An object of class fv_pcf containing the function values of the PCF and the envelope.
Details
Since the pair-correlation function is a density function, we employ the frequently used Epanechnikov kernel (Silverman 1986, Stoyan and Stoyan 1994, Nuske et al. 2009). The Epanechnikov kernel is a weight function putting maximal weight to pairs with distance exactly equal to r but also incorporating pairs only roughly at distance r with reduced weight. This weight falls to zero if the actual distance between the points differs from r by at least \(\delta\), the so-called bandwidth parameter, which determines the degree of smoothness of the function. Penttinen et al. (1992) and Stoyan and Stoyan (1994) suggest to set c aka stoyan-parameter of \(c / {\sqrt{\lambda}}\) between 0.1 and 0.2 with \(\lambda\) being the intensity of the pattern.
The edge correction is based on suggestions by Ripley (1981). For each pair of objects \(i\) and \(j\), a buffer with buffer distance \(r_{ij}\) is constructed around the object \(i\). The object \(j\) is then weighted by the inverse of the ratios \(p_{ij}\) of the buffer perimeter being within the study area. That way we account for the reduced probability of finding objects close to the edge of the study area.
The alpha level of the pointwise critical envelope is \(\alpha = \frac{n\_rank * 2}{n\_sim + 1}\) according to (Besag and Diggle 1977, Buckland 1984, Stoyan and Stoyan 1994).
References
Besag, J. and Diggle, P.J. (1977): Simple Monte Carlo tests for spatial pattern. Journal of the Royal Statistical Society. Series C (Applied Statistics), 26(3): 327–333. https://doi.org/10.2307/2346974
Buckland, S.T. (1984). Monte Carlo Confidence Intervals. Biometrics, 40(3): 811-817. https://doi.org/10.2307/2530926
Nuske, R.S., Sprauer, S. and Saborowski, J. (2009): Adapting the pair-correlation function for analysing the spatial distribution of canopy gaps. Forest Ecology and Management, 259(1): 107–116. https://doi.org/10.1016/j.foreco.2009.09.050
Penttinen A., Stoyan D., Henttonen H. M. (1992): Marked point processes in forest statistics. Forest Science, 38(4): 806–824. https://doi.org/10.1093/forestscience/38.4.806
Ripley, B.D. (1981): Spatial Statistics. John Wiley & Sons, New York. https://doi.org/10.1002/0471725218
Silverman, B.W. (1986): Density Estimation for Statistics and Data Analysis. Chapman and Hall, London.
Stoyan, D. and Stoyan, H. (1994) Fractals, random shapes and point fields: Methods of geometrical statistics. John Wiley & Sons, Chichester.
Examples
# it's advised against setting n_sim < 199
ds <- pat2dists(area=sim_area, pattern=sim_pat_reg, max_dist=25, n_sim=3)
# derive PCF and envelope
pcf <- dists2pcf(ds, r=0.2, r_max=25, stoyan=0.15, n_rank=1)