I have created fitted models for an inhomogeneous point pattern using spatstat. I have used the script below to do so. I would like to create envelopes for the fitted K-function. However, when using the envelope(), it doesn't seem to recognize the lamda parameter ##inhomogeneous functions based on top fit covariate model### lam <- fitted(invert.cord.model, location=nestsF.ppp, dataonly=TRUE) Ki <- Kinhom(nests.ppp, lam, correction="translate") plot(Ki, sqrt(./pi) - r ~ r, xlab="r(meters)", main="Inhomogeneous L-Function- Controlling for Inverts") g1 <- pcf(Ki, spar=0.9, method="b") plot(g1, main="Inhomogeneous Pair Correlation Function: Inverts") Thanks in advance!! Raya -- Raya A Pruner - Graduate Student University of Florida Department of Wildlife Ecology and Conservation PO Box 110430 Gainesville, FL 32611-0430 352-214-3262
envelopes to inhomogeneous k-function based on fitted models: with spatstat
2 messages · Raya A. Pruner, Marcelino de la Cruz
Hi Raya,
You should have provided the code that you tried
and that "doesn't seem to recognize the lamda parameter".
Anyway, I think that the problem is that you have
estimated the intensity at the data points only
(your argument dataonly=TRUE). When computing the
envelope of the Kinhom function, for each
simulated pattern you will need an estimate of
the intensity at the locations of each of the
simulated points, while your "lam" object has
only the intensity at the original point locations.
So the easiest way is using a a pixel image
(object of class "im") assumed to contain the
values of the intensity function at all locations
in the window as your lambda.
Try something like this:
lam <- predict.ppm(invert.cord.model, type="trend")
or
lam <- predict.ppm(invert.cord.model, type="trend", covariates= mycovariates)
if your model includes covariates.
Then
Ki.env <- envelope( nests.ppp, Kinhom,
lambda=lam, simulate = expression( rpoispp(lam)), correction="translate")
if your model is an heterogeneous poisson or
using other simulation function (rmh, rMaternI,
rStrauss, etc) whenever appropirate.
By the way, beware of the type of prediction
that you want from your model. By default,
fitted.ppm has argument type="lambda", which
gives you the conditional intensity ("lambda") .
I think you really need the trend, i.e. you should have asked for
lam <- fitted(invert.cord.model,
location=nestsF.ppp, dataonly=TRUE, type="trend")
Cheers,
Marcelino
At 16:37 23/03/2010, Raya A. Pruner wrote:
I have created fitted models for an inhomogeneous point pattern using spatstat. I have used the script below to do so. I would like to create envelopes for the fitted K-function. However, when using the envelope(), it doesn't seem to recognize the lamda parameter ##inhomogeneous functions based on top fit covariate model### lam <- fitted(invert.cord.model, location=nestsF.ppp, dataonly=TRUE) Ki <- Kinhom(nests.ppp, lam, correction="translate") plot(Ki, sqrt(./pi) - r ~ r, xlab="r(meters)", main="Inhomogeneous L-Function- Controlling for Inverts") g1 <- pcf(Ki, spar=0.9, method="b") plot(g1, main="Inhomogeneous Pair Correlation Function: Inverts") Thanks in advance!! Raya -- Raya A Pruner - Graduate Student University of Florida Department of Wildlife Ecology and Conservation PO Box 110430 Gainesville, FL 32611-0430 352-214-3262
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo
________________________________ Marcelino de la Cruz Rot Departamento de Biolog?a Vegetal E.U.T.I. Agr?cola Universidad Polit?cnica de Madrid 28040-Madrid Tel.: 91 336 54 35 Fax: 91 336 56 56 marcelino.delacruz at upm.es _________________________________