computeLTA {SpatioTemporal} | R Documentation |
Computes the long term average of observations and cross-validated
predictions for each of the sites in object
. The long term averages
are computed using only timepoints that have observations, this
applies to both the observed and predicted. Also the function allows for a
transformation: if requested the transformation is applied before the
averaging.
computeLTA(object, transform = function(x) { return(x) })
object |
A |
transform |
Transform observations (without bias correction) and
predictions before
computing averages; e.g. |
Returns a (number of locations) - by - 4 matrix with the observed and predicted value (using the three different model parts) for each location.
Johan Lindstrom
Other predCVSTmodel functions: estimateCV.STmodel
Other cross-validation functions: createCV
,
dropObservations
,
estimateCV.STmodel
,
predictNaive
##load data data(pred.cv.mesa) ##compute long term averages of predictions and observations pred.lta <- computeLTA(pred.cv.mesa) ##we can now compare observed and predicted averages at each site plot(pred.lta[,"obs"], pred.lta[,"EX.mu"], pch=1, xlim=range(pred.lta), ylim=range(pred.lta), xlab="obs", ylab="predictions") ##for the different model components points(pred.lta[,"obs"], pred.lta[,"EX.mu.beta"], pch=3, col=2) points(pred.lta[,"obs"], pred.lta[,"EX"], pch=4, col=3) abline(0,1) ##we could also try computaitons on the original scale pred.lta <- computeLTA(pred.cv.mesa, exp) ##compare observed and predicted averages plot(pred.lta[,"obs"], pred.lta[,"EX.mu"], pch=1, xlim=range(pred.lta), ylim=range(pred.lta), xlab="obs", ylab="predictions") points(pred.lta[,"obs"], pred.lta[,"EX.mu.beta"], pch=3, col=2) points(pred.lta[,"obs"], pred.lta[,"EX"], pch=4, col=3) abline(0,1)