Analysing spatial point patterns in R - CSIRO
Analysing spatial point patterns in R - CSIRO
Analysing spatial point patterns in R - CSIRO
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
15.3 Fitt<strong>in</strong>g Poisson processes <strong>in</strong> spatstat 97<br />
where X is the <strong>po<strong>in</strong>t</strong> pattern dataset, and ~trend is an R formula with no left-hand side. This<br />
should be viewed as a model with log l<strong>in</strong>k, so the formula ~trend specifies the form of the<br />
logarithm of the <strong>in</strong>tensity function.<br />
To fit the homogeneous Poisson model:<br />
> ppm(bei, ~1)<br />
Stationary Poisson process<br />
Uniform <strong>in</strong>tensity: 0.007208<br />
To fit an <strong>in</strong>homogeneous Poisson model with an <strong>in</strong>tensity that is log-l<strong>in</strong>ear <strong>in</strong> the cartesian<br />
coord<strong>in</strong>ates, i.e. λ θ ((x,y)) = exp(θ 0 + θ 1 x + θ 2 y),<br />
> ppm(bei, ~x + y)<br />
Nonstationary Poisson process<br />
Trend formula: ~x + y<br />
Fitted coefficients for trend formula:<br />
(Intercept) x y<br />
-4.7245290274 -0.0008031288 0.0006496090<br />
Herexandyare reserved names that always refer to the cartesian coord<strong>in</strong>ates. In the output,<br />
the ‘fitted coefficients’ are the maximum likelihood estimates of θ 0 ,θ 1 ,θ 2 , the coefficients of the<br />
‘l<strong>in</strong>ear predictor’. The fitted <strong>in</strong>tensity function is<br />
λ θ ((x,y)) = exp (−4.724529 + −0.000803x + 0.00065y) .<br />
To fit an <strong>in</strong>homogeneous Poisson model with an <strong>in</strong>tensity that is log-quadratic <strong>in</strong> the cartesian<br />
coord<strong>in</strong>ates, i.e. such that log λ θ ((x,y)) is a quadratic <strong>in</strong> x and y:<br />
> ppm(bei, ~polynom(x, y, 2))<br />
Nonstationary Poisson process<br />
Trend formula: ~polynom(x, y, 2)<br />
Fitted coefficients for trend formula:<br />
(Intercept) polynom(x, y, 2)[x] polynom(x, y, 2)[y]<br />
-4.275762e+00 -1.609187e-03 -4.895166e-03<br />
polynom(x, y, 2)[x^2] polynom(x, y, 2)[x.y] polynom(x, y, 2)[y^2]<br />
1.625968e-06 -2.836387e-06 1.331331e-05<br />
Essentially any k<strong>in</strong>d of model formula can be used, <strong>in</strong>volv<strong>in</strong>g the reserved names x and y<br />
and any covariates (as we expla<strong>in</strong> later).<br />
To fit a model with constant but unequal <strong>in</strong>tensities on each side of the vertical l<strong>in</strong>e x = 500,<br />
the explanatory variable S(u) should be a factor with two levels, Left and Right say, tak<strong>in</strong>g<br />
the value Left when the location u is to the left of the l<strong>in</strong>e x = 500.<br />
Copyright<strong>CSIRO</strong> 2010