Predict Density Surface
Predict density at each point on a raster mask from a fitted secr model.
predictDsurface(object, mask = NULL, se.D = FALSE, cl.D = FALSE, alpha =
0.05, parameter = c('D', 'noneuc'))object |
fitted secr object |
mask |
secr mask object |
se.D |
logical for whether to compute prediction SE |
cl.D |
logical for whether to compute confidence limits |
alpha |
alpha level for 100(1 – alpha)% confidence intervals |
parameter |
character for real parameter to predict |
Predictions use the linear model for density on the link scale in the
fitted secr model ‘object’, or the fitted user-defined function, if
that was specified in secr.fit.
If ‘mask’ is NULL then predictions are for the mask component of ‘object’.
SE and confidence limits are computed only if specifically requested. They are not available for user-defined density functions.
Density is adjusted automatically for the number of clusters in
‘mashed’ models (see mash).
Object of class ‘Dsurface’ inheriting from ‘mask’. Predicted densities are added to the covariate dataframe (attribute ‘covariates’) as column(s) with prefix ‘D.’ If the model uses multiple groups, multiple columns will be distinguished by the group name (e.g., "D.F" and "D.M"). If groups are not defined the column is named "D.0".
For multi-session models the value is a multi-session mask.
The pointwise prediction SE is saved as a covariate column prefixed ‘SE.’ (or multiple columns if multiple groups). Confidence limits are likewise saved with prefixes ‘lcl.’ and ‘ucl.’.
## use canned possum model
shorePossums <- predictDsurface(possum.model.Ds)
par(mar = c(1,1,1,6))
plot(shorePossums, plottype = "shaded", polycol = "blue", border = 100)
plot(traps(possumCH), detpar = list(col = "black"), add = TRUE)
par(mar = c(5,4,4,2) + 0.1) ## reset to default
## extract and summarise
summary(covariates(shorePossums))
## Not run:
## extrapolate to a new mask; add covariate needed by model; plot
regionmask <- make.mask(traps(possumCH), buffer = 1000, spacing = 10,
poly = possumremovalarea)
dts <- distancetotrap(regionmask, possumarea)
covariates(regionmask) <- data.frame(d.to.shore = dts)
regionPossums <- predictDsurface(possum.model.Ds, regionmask,
se.D = TRUE, cl.D = TRUE)
par(mfrow = c(1,2), mar = c(1,1,1,6))
plot(regionPossums, plottype = "shaded", mesh = NA, breaks = 20)
plot(regionPossums, plottype = "contour", add = TRUE)
plot(regionPossums, covariate = "SE", plottype = "shaded",
mesh = NA, breaks = 20)
plot(regionPossums, covariate = "SE", plottype = "contour",
add = TRUE)
## confidence surfaces
plot(regionPossums, covariate = "lcl", breaks = seq(0,3,0.2),
plottype = "shaded")
plot(regionPossums, covariate = "lcl", plottype = "contour",
add = TRUE, levels = seq(0,2.7,0.2))
title("lower 95% surface")
plot(regionPossums, covariate = "ucl", breaks=seq(0,3,0.2),
plottype = "shaded")
plot(regionPossums, covariate = "ucl", plottype = "contour",
add = TRUE, levels = seq(0,2.7,0.2))
title("upper 95% surface")
## annotate with CI
par(mfrow = c(1,1))
plot(regionPossums, plottype = "shaded", mesh = NA, breaks = 20)
plot(traps(possumCH), add = TRUE, detpar = list(col = "black"))
spotHeight(regionPossums, dec = 1, pre = c("lcl","ucl"), cex = 0.8)
## perspective plot
pm <- plot(regionPossums, plottype = "persp", box = FALSE, zlim =
c(0,3), phi=30, d = 5, col = "green", shade = 0.75, border = NA)
lines(trans3d (possumremovalarea$x, possumremovalarea$y,
rep(1,nrow(possumremovalarea)), pmat = pm))
par(mfrow = c(1,1), mar = c(5, 4, 4, 2) + 0.1) ## reset to default
## compare estimates of region N
## grid cell area is 0.01 ha
sum(covariates(regionPossums)[,"D.0"]) * 0.01
region.N(possum.model.Ds, regionmask)
## End(Not run)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.