Create likelihood help

BayesianTools/R/classLikelihood.R

@@ -8,6 +8,7 @@
 #' @param sampler sampler
 #' @seealso \code{\link{likelihoodIidNormal}} \cr
 #'          \code{\link{likelihoodAR1}} \cr
+#' @example /inst/examples/createLikelihoodHelp.R        
 #' @export
 createLikelihood <- function(likelihood, names = NULL, parallel = F, catchDuplicates=T, 
                              sampler = NULL, parallelOptions = NULL){

BayesianTools/R/plotSensitivityOAT.R

@@ -2,11 +2,18 @@
 #' @author Florian Hartig
 #' @param bayesianSetup An object of class BayesianSetup
 #' @param selection indices of selected parameters
-#' @param equalScale if T, y axis of all plots will have the same scale
+#' @param equalScale logical, if TRUE - y axis of all plots will have the same scale
 #' @note This function can also be used for sensitivity analysis of an arbitrary output - just create a BayesianSetup with this output. 
+#' @details
+#' When the scale of the parameter do not match then the plot looks like this :
+#' ![](plotSensitivity-ScaleDontMatch.png "Sensitivity Plot")
+#' 
+#' In that case, we should put 'equalScale = F'. Then the plot looks like this:
+#' ![](plotSensitivity-ScaleFalse.png "Sensitivity Plot")
+#' 
 #' @example /inst/examples/plotSensitivityHelp.R
 #' @export
-plotSensitivity <- function(bayesianSetup, selection = NULL, equalScale = T){
+plotSensitivity <- function(bayesianSetup, selection = NULL, equalScale = TRUE){
 
   if (is.null(selection)) selection = 1:bayesianSetup$numPars
   n = length(selection)
@@ -39,7 +46,7 @@ plotSensitivity <- function(bayesianSetup, selection = NULL, equalScale = T){
 
 
   for (i in 1:n){
-    if(equalScale == T) plot(resp~par, xlab = names[i], type = "l", col = "red", data = post[[i]], ylim = c(minR, maxR), ylab = "Response")
+    if(equalScale == TRUE) plot(resp~par, xlab = names[i], type = "l", col = "red", data = post[[i]], ylim = c(minR, maxR), ylab = "Response")
     else plot(resp~par, xlab = names[i], type = "l", col = "red", data = post[[i]], ylab = "Response")
 
     abline(v = refPar[i])

BayesianTools/inst/examples/createLikelihoodHelp.R

@@ -0,0 +1,63 @@
+
+
+data = rnorm(20)
+
+# create standardized likelihood density for Gaussian likelihood function
+# fitting parameters mean and standard deviation
+
+likelihood <- function(x) {
+  predicted <- rep(x[1], length(data))
+  LL = likelihoodIidNormal(predicted, data, x[2])
+  return(LL)
+}
+
+likelihood(x = c(1,1))
+
+# The next step is optional, you can provide the likelihood density directly
+# to BayesianSetup and createLikelihood will be called automatically.
+# The purpose of the createLikelihood function is to parallelize the 
+# density function and to add some convenience functions, 
+# for example to catch exceptions that are raised by the call to likelihood
+likelihoodClass <- createLikelihood(likelihood)
+
+bayesianSetup <- createBayesianSetup(likelihood = likelihoodClass, 
+                                     lower = c(-10,0.001), 
+                                     upper = c(10, 5))
+
+settings = list(iterations = 1000)
+out <- runMCMC(bayesianSetup, sampler = "DEzs", settings)
+summary(out)
+
+
+
+# simulate temporally autocorrelated data
+AR1sim<-function(n, a){
+  x = rep(NA, n)
+  x[1] = 0
+  for(i in 2:n){
+    x[i] = a * x[i-1] + (1-a) * rnorm(1)
+  }
+  return(x)
+}
+
+set.seed(123)
+data = AR1sim(1000, 0.5)
+plot(data, type = "l") 
+
+
+# create likelihood function for AR1 likelihood function
+likelihood <- function(x) {
+  predicted <- rep(x[1], length(data))
+  LL = likelihoodAR1(predicted, data, x[2], x[3])
+  return(LL)
+}
+
+bayesianSetup <- createBayesianSetup(likelihood = likelihood, 
+                                     lower = c(-10,0.001, 0), 
+                                     upper = c(10, 5, 0.99))
+
+settings = list(iterations = 1000)
+out <- runMCMC(bayesianSetup, sampler = "DEzs", settings)
+summary(out, start = 200)
+plot(out, start = 200)
+

BayesianTools/inst/examples/plotSensitivityHelp.R

@@ -1,5 +1,22 @@
+data = rnorm(20)
+
+# create standardized likelihood density for Gaussian likelihood function
+# fitting parameters mean and standard deviation
+likelihood <- function(x) {
+  predicted <- rep(x[1], length(data))
+  LL = likelihoodIidNormal(predicted, data, x[2])
+  return(LL)
+}
+
+setup = createBayesianSetup(likelihood, lower = c(-10, 0.01), upper = c(10,5))
+
+# showing posterior response surface for parameter 1
+# note that this is plotted at 
+setup$prior$best
+plotSensitivity(setup, selection = 2)
+
+plotSensitivity(setup)
+# when parameters are not on the same scale, we can use
+plotSensitivity(setup, equalScale = FALSE)
 
-ll <- testDensityBanana
-bayesianSetup <- createBayesianSetup(likelihood = ll, lower = rep(-10, 2), upper = rep(10, 2))
 
-plotSensitivity(bayesianSetup)

Development/Figures/figures.R

@@ -0,0 +1,21 @@
+##### plotSensitivity #####
+
+data = rnorm(20)
+
+# create standardized likelihood density for Gaussian likelihood function
+# fitting parameters mean and standard deviation
+likelihood <- function(x) {
+  predicted <- rep(x[1], length(data))
+  LL = likelihoodIidNormal(predicted, data, x[2])
+  return(LL)
+}
+
+setup = createBayesianSetup(likelihood, lower = c(-10, 0.01), upper = c(10,5))
+
+png(filename = "BayesianTools/man/figures/plotSensitivity-ScaleDontMatch.png", width = 400, height = 300)
+plotSensitivity(setup)
+dev.off()
+
+png(filename = "BayesianTools/man/figures/plotSensitivity-ScaleFalse.png", width = 400, height = 300)
+plotSensitivity(setup, equalScale = F)
+dev.off()