Effects names and exponentiation of the effects updated.

R/clmplusAggregateDataPP.R

@@ -31,6 +31,7 @@
 #'   age terms, \code{"1"} for \eqn{\beta_x^{(0)}=1}, a predefined parametric 
 #'   function of age (see details) or \code{NULL} if there is no cohort effect. 
 #'   To be disregarded unless the practitioner specifies his own hazard model in StMoMo. 
+#' @param effect_log_scale \code{logical}, whether effects should be on the logarithmic scale. By default, \code{TRUE}.
 #' @param constFun \code{function}, it defines the identifiability constraints of the 
 #'   model. It must be a function of the form 
 #'   \code{constFun <- function(ax, bx, kt, b0x, gc, wxt, ages)} taking a set
@@ -59,6 +60,7 @@ clmplus <- function(AggregateDataPP,
                     staticAgeFun = TRUE, 
                     periodAgeFun = "NP",
                     cohortAgeFun = NULL, 
+                    effect_log_scale=TRUE,
                     constFun = function(ax, bx, kt, b0x, gc, wxt, ages) list(ax = ax, bx = bx, kt = kt, b0x = b0x, gc = gc),
                     ...){
 
@@ -96,6 +98,7 @@ clmplus <- function(AggregateDataPP,
 #'   age terms, \code{"1"} for \eqn{\beta_x^{(0)}=1}, a predefined parametric 
 #'   function of age (see details) or \code{NULL} if there is no cohort effect. 
 #'   To be disregarded unless the practitioner specifies his own hazard model in StMoMo. 
+#' @param effect_log_scale \code{logical}, whether effects should be on the logarithmic scale. By default, \code{TRUE}.
 #' @param constFun \code{function}, it defines the identifiability constraints of the 
 #'   model. It must be a function of the form 
 #'   \code{constFun <- function(ax, bx, kt, b0x, gc, wxt, ages)} taking a set
@@ -122,6 +125,7 @@ clmplus.default <- function(AggregateDataPP,
                             staticAgeFun = TRUE, 
                             periodAgeFun = "NP",
                             cohortAgeFun = NULL, 
+                            effect_log_scale=TRUE,
                             constFun = function(ax, bx, kt, b0x, gc, wxt, ages) list(ax = ax, bx = bx, kt = kt, b0x = b0x, gc = gc),
                             ...){message('The object provided must be of class AggregateDataPP')}
 
@@ -157,6 +161,7 @@ clmplus.default <- function(AggregateDataPP,
 #'   age terms, \code{"1"} for \eqn{\beta_x^{(0)}=1}, a predefined parametric 
 #'   function of age (see details) or \code{NULL} if there is no cohort effect. 
 #'   To be disregarded unless the practitioner specifies his own hazard model in StMoMo. 
+#' @param effect_log_scale \code{logical}, whether effects should be on the logarithmic scale. By default, \code{TRUE}.
 #' @param constFun \code{function}, it defines the identifiability constraints of the 
 #'   model. It must be a function of the form 
 #'   \code{constFun <- function(ax, bx, kt, b0x, gc, wxt, ages)} taking a set
@@ -197,6 +202,7 @@ clmplus.AggregateDataPP <- function(AggregateDataPP,
                             staticAgeFun = TRUE, 
                             periodAgeFun = "NP",
                             cohortAgeFun = NULL, 
+                            effect_log_scale=TRUE,
                             constFun = function(ax, bx, kt, b0x, gc, wxt, ages) list(ax = ax, bx = bx, kt = kt, b0x = b0x, gc = gc),
                             ...){
 
@@ -278,7 +284,8 @@ clmplus.AggregateDataPP <- function(AggregateDataPP,
   fitted_effects <- pkg.env$find.fitted.effects(J,
                                                 age.eff= model$ax,
                                                 cohort.eff= model$gc,
-                                                period.eff=model$kt)
+                                                period.eff=model$kt,
+                                                effect_log_scale=effect_log_scale)
 
   }
 

R/globalVariables.R

@@ -672,12 +672,13 @@ pkg.env$find.development.factors <- function(J,
 pkg.env$find.fitted.effects <- function(J,
                                        age.eff,
                                        period.eff,
-                                       cohort.eff){
+                                       cohort.eff,
+                                       effect_log_scale){
 
   # Function that finds the fitted effects.
 
   ax <- c(NA,age.eff[!is.na(age.eff)])
-  names(ax) <- c(1:length(ax))
+  names(ax) <- c(0:(length(ax)-1))
 
   if(!is.null(cohort.eff)){
     gc <-  c(cohort.eff[!is.na(cohort.eff)],NA) #last cohort effect will be extrapolated
@@ -688,7 +689,7 @@ pkg.env$find.fitted.effects <- function(J,
   if(!is.null(period.eff)){
 
     kt <-  c(NA,period.eff[!is.na(period.eff)]) #the first period is disregarded
-    names(kt) <- c(1:length(kt))
+    names(kt) <- c(0:(length(kt)-1))
 
   }else{kt<-NULL}
 
@@ -698,6 +699,8 @@ pkg.env$find.fitted.effects <- function(J,
     fitted_accident_effect= gc
     )
 
+  if(effect_log_scale==FALSE){out<-lapply(out,exp)}
+
   return(out)
 
 

README.md

@@ -65,6 +65,7 @@ datapp = AggregateDataPP(cumulative.payments.triangle = dataset, eta= 1/2)
 ```
 a.model.fit=clmplus(datapp,
                  hazard.model = "a") # age-model replicates the chain ladder
+                 
 ac.model.fit=clmplus(datapp,
                  hazard.model = "ac")
 
@@ -85,21 +86,28 @@ plot(apc.model.fit)
 Predictions are performed with the `predict` function.
 
 ```
+a.model=predict(a.model.fit)
+                 
+# clmplus reserve (age model)
+sum(a.model$reserve)
+#226875.5
+
+
 ac.model=predict(ac.model.fit,
                  gk.fc.model = 'a',
                  gk.order = c(1,1,0))
                  
 # clmplus reserve (age-cohort model)
 sum(ac.model$reserve)
-#38126.05
+#205305.7
 
 ap.model<- predict(ap.model.fit,
                  ckj.fc.model = 'a',
                  ckj.order = c(0,1,0))
 
 # clmplus reserve (age-period model)
 sum(ap.model$reserve)
-#37375.01
+#215602.8
           
                  
 apc.model<-predict(apc.model.fit,
@@ -109,7 +117,7 @@ apc.model<-predict(apc.model.fit,
                   ckj.order = c(0,1,0))
 # clmplus reserve (age-period-cohort model)
 sum(apc.model$reserve)
-#38498.54
+#213821.6
 ```
 
 The fitted effect (and extrapolated) effects can be inspected with the `plot` function. We continue below the example with the `apc` model.