roiKrige: Prediction at new location using local regression and...
In martindurocher/floodRFA: A R-package for regional frequency of floods

Description Usage Arguments Value See Also Examples

Return the prediction of a local regression model using region of influence (ROI). The residuals are further predicted using kriging or thin plate spline.

1 2	roiKrige(form, x, xnew, nk, loc, grp = NULL, model = "none", loc.spatial = NULL, ker = TRUE)

`form`	Formula defining the trend.
`x`	Data for trainging the model.
`xnew`	Data at new locations. Validation set.
`nk`	Size of the neighborhoods.
`loc`	Formula. Covariates of the similarity between sites (euclidean distance). Used for delineation.
`grp`	Factor defining groups for calibrating `nk`.
`model`	Model for predicting the residuals. Could be 'tps' or a vector of variogram models. See vgm.
`loc.spatial`	Formula defining the spatial covariates (euclidean distance). Used for predicting residuals with spatial correlation.
`ker`	Logical, should weight be used to favor closer sites.
`...`

`pred`	Prediction at new sites.
`pred.se`	Standard deviation at new sites.
`trend`	Part of the prediction attributed to trend
`trend.se`	Standard deviation associated with the trend
`fitted`	Fitted values (training sites)
`fitted.se`	Standard deviaton of the fitted values
`vgm`	Sample variogram.
`model`	Fitted variogram model. See vgm.

gamKrige, nnKrige

 # Gather info in on data.frame
 xd <- cbind( l1 = log(sapply(floodStream, mean)),
             floodVars, lon = floodCoord[,1], lat = floodCoord[,2])
 nsite <- nrow(xd)

 # identify a validation and a training set
 valid <- seq(nsite) %in% sample(seq(nsite), round(.2*nsite))
 train <- !valid

 l1Form <- l1 ~ area + slope + elev + map + len + wb
 nform <- ~ area + slope + elev + map + len + wb

 # Using kriging
 fit <- roiKrige(l1Form, x = xd[train,], xnew = xd[valid,],
                 n = 50, ker = TRUE,
                model = c('Exp','Gau','Sph'),
                loc = nform, loc.spatial = ~lon+lat)
 print(fit)
 plot(fit)
 predict(fit)

 # Without spatial component
 fit <- roiKrige(l1Form, x = xd[train,], xnew = xd[valid,],
                 n = 20, ker = FALSE, model = 'none', loc = nform)

 # Using Thin plate spline
 fit <- roiKrige(l1Form, x = xd[train,], xnew = xd[valid,],
                 n = 20, loc = nform, model = 'tps',
                loc.spatial = ~lon+lat)