Commit d4f4a873 authored by Delaigue Olivier's avatar Delaigue Olivier
Browse files

v1.2.14.14 NEW: box-cox transformation allowed in CreateInputsCrit and computed in .ErrorCrit

parent 3565d8b5
Package: airGR
Type: Package
Title: Suite of GR Hydrological Models for Precipitation-Runoff Modelling
Version: 1.2.14.12
Version: 1.2.14.14
Date: 2019-04-17
Authors@R: c(
person("Laurent", "Coron", role = c("aut", "trl"), comment = c(ORCID = "0000-0002-1503-6204")),
......
......@@ -14,17 +14,17 @@ output:
### 1.2.14.12 Release Notes (2019-04-17)
### 1.2.14.14 Release Notes (2019-04-17)
#### New features
- <code>CreateInputsCrit()</code> now allows power transformation (as numeric or character values) in the <code>transfo</code> argument.
- <code>CreateInputsCrit()</code> now allows power (as numeric or character values) and the Box-Cox transformations in the <code>transfo</code> argument.
#### Minor user-visible changes
- <code>.ErrorCrit()</code> private function added to check inputs into <code>ErrorCrit_&#42;()</code> functions.
- <code>.ErrorCrit()</code> private function added to check inputs into <code>ErrorCrit_&#42;()</code> functions. The <code>ErrorCrit_&#42;()</code> functions were simplified accordingly.
____________________________________________________________________________________
......
......@@ -157,7 +157,7 @@ CreateInputsCrit <- function(FUN_CRIT,
inVarObs <- c("Q", "SCA", "SWE")
msgVarObs <- "'VarObs' must be a (list of) character vector(s) and one of %s"
msgVarObs <- sprintf(msgVarObs, paste(sapply(inVarObs, shQuote), collapse = ", "))
inTransfo <- c("", "sqrt", "log", "inv", "sort") # pow is not checked by inTransfo, but appears in the warning message and checkef after (see ## check 'transfo')
inTransfo <- c("", "sqrt", "log", "inv", "sort", "boxcox") # pow is not checked by inTransfo, but appears in the warning message and checkef after (see ## check 'transfo')
msgTransfo <- "'transfo' must be a (list of) character vector(s) and one of %s, or numeric value for power transformation"
msgTransfo <- sprintf(msgTransfo, paste(sapply(inTransfo, shQuote), collapse = ", "))
......
......@@ -102,7 +102,7 @@
if (InputsCrit$transfo == "") {
CritName <- paste0(Crit, "[CritVar]")
}
if (InputsCrit$transfo %in% c("sqrt", "log", "sort")) {
if (InputsCrit$transfo %in% c("sqrt", "log", "sort", "boxcox")) {
CritName <- paste0(Crit, "[", InputsCrit$transfo, "(CritVar)]")
}
if (InputsCrit$transfo == "inv") {
......@@ -172,6 +172,10 @@
VarObs <- sort(VarObs, na.last = TRUE)
InputsCrit$BoolCrit <- sort(InputsCrit$BoolCrit, decreasing = TRUE)
}
if (InputsCrit$transfo == "boxcox") {
VarSim <- (VarSim^0.25 - 0.01 * mean(VarSim, na.rm = TRUE)) / 0.25
VarObs <- (VarObs^0.25 - 0.01 * mean(VarObs, na.rm = TRUE)) / 0.25
}
if (grepl("\\^", InputsCrit$transfo)) {
VarObs <- VarObs^transfoPow
VarSim <- VarSim^transfoPow
......
......@@ -32,7 +32,7 @@ CreateInputsCrit(FUN_CRIT, InputsModel, RunOptions,
\item{BoolCrit}{(optional) [boolean (atomic or list)] boolean (the same length as \code{Obs}) giving the time steps to consider in the computation (all time steps are considered by default)}
\item{transfo}{(optional) [character (atomic or list)] name of the transformation (e.g. \code{""}, \code{"sqrt"}, \code{"log"}, \code{"inv"}, \code{"sort"} or numeric value for power transformation (see details))}
\item{transfo}{(optional) [character (atomic or list)] name of the transformation (e.g. \code{""}, \code{"sqrt"}, \code{"log"}, \code{"inv"}, \code{"sort"}, \code{"boxcox"} or a numeric value for power transformation (see details))}
\item{Weights}{(optional) [numeric (atomic or list)] vector of weights necessary to calculate a composite criterion (the same length as \code{FUN_CRIT}) giving the weights to use for elements of \code{FUN_CRIT} [-]. See details}
......@@ -53,7 +53,7 @@ CreateInputsCrit(FUN_CRIT, InputsModel, RunOptions,
\emph{$Obs } \tab [numeric] series of observed variable(s) ([mm/time step] for discharge or SWE, [-] for SCA) \cr
\emph{$VarObs } \tab [character] names of the observed variable(s) \cr
\emph{$BoolCrit } \tab [boolean] boolean giving the time steps considered in the computation \cr
\emph{$transfo } \tab [character] name of the transformation (e.g. \code{""}, \code{"sqrt"}, \code{"log"}, \code{"inv"}, \code{"sort"}) \cr
\emph{$transfo } \tab [character] name of the transformation (e.g. \code{""}, \code{"sqrt"}, \code{"log"}, \code{"inv"}, \code{"sort"}, \code{"boxcox"} or a number for power transformation) \cr
\emph{$epsilon } \tab [numeric] small value to add to all observations and simulations when \code{"log"} or \code{"inv"} transformations are used [same unit as \code{Obs}] \cr
\emph{$Weights } \tab [numeric] vector (same length as \code{VarObs}) giving the weights to use for elements of \code{FUN_CRIT} [-] \cr
}
......@@ -72,14 +72,16 @@ Creation of the \code{InputsCrit} object required to the \code{ErrorCrit_*} func
\details{
Users wanting to use \code{FUN_CRIT} functions that are not included in the package must create their own InputsCrit object accordingly. \cr \cr
The syntax of the power transformation can be written in numerical form or as a string of characters, for example for a square: \code{transfo = 2}, \code{"2"} or \code{"^2"}.
The syntax of the power transformation allows a numeric or a string of characters. For example for a squared transformation, the following can be used: \code{transfo = 2}, \code{transfo = "2"} or \code{transfo = "^2"}. Negative values are allowed. Fraction values are not allowed (e.g., \code{"-1/2"} must instead be written \code{"-0.5"}).
\cr \cr
In order to make sure that KGE and KGE2 remain dimensionless and not impacted by zero values, the Box-Cox transformation (\code{transfo = "boxcox"}) uses the formulation given in Equation 10 of Santos et al. (2018). Lambda is set to 0.25 accordingly.
\cr \cr
The epsilon value is useful when \code{"log"} or \code{"inv"} transformations are used (to avoid calculation of the inverse or of the logarithm of zero). The impact of this value and a recommendation about the epsilon value to use (usually one hundredth of average observation) are discussed in Pushpalatha et al. (2012) for NSE and in Santos et al. (2018) for KGE and KGE'. \cr \cr
We do not advise computing KGE or KGE' with log-transformation as it might be wrongly influenced by discharge values close to 0 or 1 and the criterion value is dependent on the discharge unit. See Santos et al. (2018) for more details and alternative solutions (see the references list below). \cr \cr
Users can set the following arguments as atomic or list: \code{FUN_CRIT}, \code{Obs}, \code{VarObs}, \code{BoolCrit}, \code{transfo}, \code{Weights}. If the list format is chosen, all the lists must have the same length. \cr
Calculation of a single criterion (e.g. NSE computed on discharge) is prepared by providing to \code{CreateInputsCrit} arguments atomics only. \cr
Calculation of multiple criteria (e.g. NSE computed on discharge and RMSE computed on discharge) is prepared by providing to \code{CreateInputsCrit} arguments lists except for \code{Weights} that must be set as \code{NULL}. \cr
Calculation of a composite criterion (e.g. the average between NSE computed on dscharge and NSE computed on log of discharge) is prepared by providing to \code{CreateInputsCrit} arguments lists including \code{Weights}. \cr
Calculation of a composite criterion (e.g. the average between NSE computed on discharge and NSE computed on log of discharge) is prepared by providing to \code{CreateInputsCrit} arguments lists including \code{Weights}. \cr
\code{\link{ErrorCrit_RMSE}} cannot be used in a composite criterion since it is not a unitless value.
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment