Synchronize Feature Branch

Ropenbt/DESCRIPTION

@@ -7,7 +7,7 @@ Authors@R: c(person('Matthew', 'Pratola', role=c('aut','cre','cph'),email='mprat
 URL: http://www.github.com/bandframework/OpenBT
 Description: 
     This R package provides an R interface to the OpenBT library of Bayesian Tree models (https://www.github.com/bandframework/OpenBT).  For more information and examples, refer to the README at the root of the OpenBT repository.
-License: MIT
+License: MIT + file LICENSE
 Depends: R (>= 2.10)
 Imports: data.table, zip, emoa, Hmisc
 Suggests: 

Ropenbt/LICENSE

@@ -0,0 +1 @@
+../LICENSE

Ropenbt/NAMESPACE

@@ -1,5 +1,5 @@
 export(openbt)
-S3method(predict, openbt)
+export(predict.openbt)
 export(vartivity.openbt)
 export(sobol.openbt)
 export(mopareto.openbt)
@@ -9,6 +9,18 @@ export(makecuts.openbt)
 export(setvarcuts.openbt)
 export(calcMBD)
 export(calcRS)
+S3method(plot, OpenBT_vartivity)
+S3method(print, OpenBT_posterior)
+S3method(summary, OpenBT_posterior)
+S3method(print, OpenBT_predict)
+S3method(summary, OpenBT_predict)
+S3method(print, OpenBT_sobol)
+S3method(summary, OpenBT_sobol)
+S3method(plot, OpenBT_sobol)
+S3method(print, OpenBT_vartivity)
+S3method(summary, OpenBT_vartivity)
+S3method(summary, OpenBT_cutinfo)
+S3method(print, OpenBT_cutinfo)
 importFrom(stats,cor)
 importFrom("graphics", "boxplot", "par")
 importFrom("graphics", "abline", "lines", "plot", "points")

Ropenbt/R/openbt.R

@@ -653,7 +653,7 @@ if(fit$modeltype==MODEL_MIXBART){
   res$m.upper=apply(res$mdraws,2,quantile,q.upper)
   res$smean=apply(res$sdraws,2,mean)
   res$ssd=apply(res$sdraws,2,sd)
-  res$s.5=apply(res$sdraws,2,median)
+  res$s.5=apply(res$sdraws,2,quantile,0.5)
   res$s.lower=apply(res$sdraws,2,quantile,q.lower)
   res$s.upper=apply(res$sdraws,2,quantile,q.lower)
   res$pdraws=NULL
@@ -2218,8 +2218,9 @@ loadRData <- function(fname)
 }
 
 
-print.OpenBT_posterior<-function(post)
+print.OpenBT_posterior<-function(x, ...)
 {
+   post = x
    MODEL_BT=1
    MODEL_BINOMIAL=2
    MODEL_POISSON=3
@@ -2262,13 +2263,15 @@ print.OpenBT_posterior<-function(post)
    summary(post$xicuts)
 }
 
-summary.OpenBT_posterior<-function(post)
+summary.OpenBT_posterior<-function(object, ...)
 {
+   post = object
    cat("No summary method for object.\n")
 }
 
-print.OpenBT_predict<-function(pred)
+print.OpenBT_predict<-function(x, ...)
 {
+   pred = x
    MODEL_BT=1
    MODEL_BINOMIAL=2
    MODEL_POISSON=3
@@ -2289,20 +2292,23 @@ print.OpenBT_predict<-function(pred)
    cat("Variance quantiles: ",pred$q.lower,",",pred$q.upper,"\n\n")
 }
 
-summary.OpenBT_predict<-function(pred)
+summary.OpenBT_predict<-function(object, ...)
 {
+   pred = object
    cat("No summary method for object.\n")
 }
 
-print.OpenBT_sobol<-function(sobol)
+print.OpenBT_sobol<-function(x, ...)
 {
+   sobol = x
    cat("OpenBT Sobol Indices\n")
    cat(ncol(sobol$vidraws), " variables.\n")
    cat(nrow(sobol$vidraws), " realizations.\n")
 }
 
-summary.OpenBT_sobol<-function(sobol)
+summary.OpenBT_sobol<-function(object, ...)
 {
+   sobol = object
    cat("Summary of Posterior Sobol Sensitivity Indices\n")
 
    cat("Expected Sobol Indices (Mean)\n")
@@ -2316,23 +2322,26 @@ summary.OpenBT_sobol<-function(sobol)
    print(sobol$sij.sd)
 }
 
-plot.OpenBT_sobol<-function(sobol)
+plot.OpenBT_sobol<-function(x, ...)
 {
+   sobol = x
    par(mfrow=c(3,1))
    boxplot(sobol$sidraws,ylab="Sobol Sensitivity",main="First Order Sobol Indices",xlab="Variables")
    boxplot(sobol$sijdraws,ylab="Sobol Sensitivity",main="Two-way Sobol Indices",xlab="Variables")
    boxplot(sobol$tsidraws,ylab="Sobol Sensitivity",main="Total Sobol Indices",xlab="Variables")
 }
 
-print.OpenBT_vartivity<-function(vartivity)
+print.OpenBT_vartivity<-function(x, ...)
 {
+   vartivity = x
    cat("OpenBT Variable Activity\n")
    cat(ncol(vartivity$vdraws), " variables.\n")
    cat(nrow(vartivity$vdraws), " realizations.\n")
 }
 
-summary.OpenBT_vartivity<-function(vartivity)
+summary.OpenBT_vartivity<-function(object, ...)
 {
+   vartivity = object
    cat("Summary of Posterior Variable Activity\n")
 
    p=length(vartivity$mvdraws)
@@ -2374,16 +2383,18 @@ summary.OpenBT_vartivity<-function(vartivity)
    }
 }
 
-plot.OpenBT_vartivity<-function(vartivity)
+plot.OpenBT_vartivity<-function(x, ...)
 {
+   vartivity = x
    par(mfrow=c(1,2))
    yrange=c(0,max(max(vartivity$vdraws),max(vartivity$vdrawsh)))
    boxplot(vartivity$vdraws,ylab="% node splits",main="Mean Trees",xlab="Variables",ylim=yrange)
    boxplot(vartivity$vdrawsh,ylab="% node splits",main="Variance Trees",xlab="Variables",ylim=yrange)
 }
 
-summary.OpenBT_cutinfo<-function(xi)
+summary.OpenBT_cutinfo<-function(object, ...)
 {
+   xi = object
    p=length(xi)
    cat("Number of variables: ",p,"\n")
    cat("Number of cutpoints per variable\n")
@@ -2393,8 +2404,9 @@ summary.OpenBT_cutinfo<-function(xi)
    }
 }
 
-print.OpenBT_cutinfo<-function(xi)
+print.OpenBT_cutinfo<-function(x, ...)
 {
+   xi = x
    summary.OpenBT_cutinfo(xi)
 }
 
@@ -2545,7 +2557,7 @@ mixingwts.openbt = function(
     #Now populate the summary stat matrices -- n X k matrices
     mean_matrix[,j] = apply(wt_list[[j]], 2, mean)
     sd_matrix[,j] = apply(wt_list[[j]], 2, sd)
-    med_matrix[,j] = apply(wt_list[[j]], 2, median)
+    med_matrix[,j] = apply(wt_list[[j]], 2, quantile,0.5)
     lb_matrix[,j] = apply(wt_list[[j]], 2, quantile,q.lower)
     ub_matrix[,j] = apply(wt_list[[j]], 2, quantile,q.upper)
   }

Ropenbt/man/openbt.Rd

@@ -6,8 +6,9 @@
   The function \code{openbt()} is the main function for fitting Bayesian Regression Tree models that are provided in the OpenBT project.  
 }
 \usage{
-openbt(x.train, y.train, ntree=NULL, ntreeh=NULL, ndpost=1000, 
-     nskip=100, k=NULL, power=2.0, base=.95, tc=2, sigmav=rep(1,length(y.train)), 
+openbt(x.train, y.train, f.train = matrix(1, nrow=length(y.train), ncol = 2), ntree=NULL,
+     ntreeh=NULL, ndpost=1000, nskip=100, k=NULL, power=2.0, base=.95, tc=2, 
+     sigmav=rep(1,length(y.train)), f.sd.train = NULL, wts.prior.info = NULL,
      fmean=mean(y.train), overallsd = NULL, overallnu=NULL, 
      chv = cor(x.train,method="spearman"), pbd=.7, pb=.5, stepwpert=.1, probchv=.1,
      minnumbot=5, printevery=100, numcut=100, xicuts=NULL, nadapt=1000, adaptevery=100,
@@ -16,6 +17,7 @@ openbt(x.train, y.train, ntree=NULL, ntreeh=NULL, ndpost=1000,
 \arguments{
   \item{x.train}{nxp matrix of predictor variables for the training data.}
   \item{y.train}{nx1 vector of the observed response for the training data.}
+  \item{f.train}{For eventual model mixing support.}
   \item{ntree}{Number of trees used for the mean model.}
   \item{ntreeh}{Number of trees used for the variance model.}
   \item{ndpost}{Number of iterations to run the MCMC algorithm after burn-in.}
@@ -25,6 +27,8 @@ openbt(x.train, y.train, ntree=NULL, ntreeh=NULL, ndpost=1000,
   \item{base}{Base parameter in the tree depth penalizing prior.}
   \item{tc}{Number of OpenMP threads to use.}
   \item{sigmav}{Initialization of square-root of variance parameter.}
+  \item{f.sd.train}{For eventual model mixing support.}
+  \item{wts.prior.info}{For eventual model mixing support.}
   \item{fmean}{Overall mean of the data for pre-centering the data before running the model.}
   \item{overallsd}{A rudimentary estimate of the process standard deviation. Used in calibrating the variance prior.}
   \item{overallnu}{Shape parameter for the variance prior.}

Ropenbt/man/predict.openbt.Rd

@@ -7,8 +7,9 @@
 }
 \usage{
 \method{predict}{openbt}(
-object,
+object=NULL,
 x.test=NULL,
+f.test=matrix(1, nrow = 1, ncol = 2),
 tc=2,
 fmean=fit$fmean,
 q.lower=0.025,
@@ -17,6 +18,7 @@ q.upper=0.975, ...)
 \arguments{
   \item{object}{Object of type \code{OpenBT_posterior} from a previous call to \code{openbt()}}
   \item{x.test}{New input settings in the form of an \code{npred x p} matrix at which to construct predictions.  Defaults to the training inputs.}
+  \item{f.test}{For eventual model mixing support.}
   \item{tc}{Number of CPU cores to use for parallel computing.}
   \item{fmean}{Mean-centering vector for the training data.  Defaults to the value used by \code{openbt()} when fitting the model.  Usually should be left to the default.}
   \item{q.lower}{Lower quantile to return.}