version 1.5

DESCRIPTION

@@ -1,8 +1,8 @@
 Package: polyapost
-Version: 1.4-3
-Date: 2017-03-14
+Version: 1.5
+Date: 2017-06-12
 Imports: boot, stats
-Depends: R (>= 3.0.2), rcdd (>= 1.1-10)
+Depends: R (>= 3.0.2), rcdd (>= 1.2)
 Title: Simulating from the Polya Posterior
 Author: Glen Meeden <glen@stat.umn.edu> and Radu Lazar
         <lazar@stat.umn.edu> and Charles J. Geyer <charlie@stat.umn.edu>
@@ -14,6 +14,6 @@ Description: Simulate via Markov chain Monte Carlo (hit-and-run algorithm)
     that is a subset of the unit simplex).
 License: GPL (>= 2)
 NeedsCompilation: yes
-Packaged: 2017-03-14 21:57:45 UTC; geyer
+Packaged: 2017-06-12 21:25:36 UTC; geyer
 Repository: CRAN
-Date/Publication: 2017-03-15 05:25:34
+Date/Publication: 2017-06-12 23:09:46 UTC

MD5

@@ -1,4 +1,4 @@
-d0c7287d45fe04dd2fefe24de4b372f0 *DESCRIPTION
+4ad5c38bed4845c12bcbe3ee313ee992 *DESCRIPTION
 b7608e4bdf9470a5d33898041a55e8e3 *NAMESPACE
 af205ca09fab4fe4f460b9c4348eb85d *R/checkconstr.R
 a24b9132c8639e5deaee307bc7fe9e23 *R/constrmat.R
@@ -7,18 +7,18 @@ a24b9132c8639e5deaee307bc7fe9e23 *R/constrmat.R
 05b5b6f481312e763815463282aac5da *R/cwpolya.R
 600fb1c902b0c084e6ccd9aaeea3ff7a *R/feasible.R
 c6bdabbb7c3e74638991f0aa7d0f8988 *R/feasible1.R
-abce94d7aeaad527b76013a71b6a2340 *R/hitrun.R
+0c6c524c1b385fd19e49f9ffe4559985 *R/hitrun.R
 945cc6bf2ed3409bc123f02f5bda1479 *R/means.R
 8d8ddb268afce5ec9658b67bf362220f *R/nullspace.R
 bde8037ae45667692d2a889433a73e31 *R/polyaest.R
 a1a00e355ceb163ef53365bfca339836 *R/polyap.R
 2be02f772609a4f9913e4bf2a30c876d *R/probvect.R
 7901d2bcd10051987b69652cb5a12e05 *R/probvect1.R
 be39dcf61a777387307071ea97d954aa *R/wtpolyap.R
-557a6ad5428303216e760138b7423627 *build/vignette.rds
+98df023f00dcc04d7d806e66a708ef82 *build/vignette.rds
 bb237297199405f4d9a96542f83ca996 *inst/doc/pp1.R
-a8f3f15f82ab29b69d5b2eca91871b09 *inst/doc/pp1.Rnw
-990a9da7208925d86e2c081313003c69 *inst/doc/pp1.pdf
+59ed7d248384662c24bf6c51f960546a *inst/doc/pp1.Rnw
+7d34d7af65b4753cf58a1ca7aa4546ea *inst/doc/pp1.pdf
 3b716132aa2bd8685f02914ffd67b2c1 *man/constrppmn.Rd
 dd16907048de6ac08d118282acb818ee *man/constrppprob.Rd
 9a9e1dfb73af54d6522d3901e9c7d48a *man/feasible.Rd
@@ -32,8 +32,8 @@ c0241c04b85b4ee2f65eab984c95d353 *src/Makevars
 8de2f794828c82d8e868a9dc861a0e7d *src/means.c
 d09330139cdab8b048fab14f91fe61e2 *src/polyapost.h
 949b366c817a7df2fc83d142efe9d62d *src/probvect.c
-6c8b99f22810226994b882e98443efe3 *tests/hitrun.R
-8bc7f4c14f434edd4175ead3b450ddd0 *tests/hitrun.Rout.save
+2b237a784a0c916c57e1c61331532b59 *tests/hitrun.R
+9dfc01a0b1bf362f4b9a4cb1450cbd7b *tests/hitrun.Rout.save
 5a9003046376ca7c471baa9bcdf926e7 *vignettes/charlie.bst
-a8f3f15f82ab29b69d5b2eca91871b09 *vignettes/pp1.Rnw
-20cd4814af718d87e8cc26ef7f292a6c *vignettes/ref.bib
+59ed7d248384662c24bf6c51f960546a *vignettes/pp1.Rnw
+27837cd23ede661b938ee0531693896c *vignettes/ref.bib

R/hitrun.R

@@ -8,8 +8,8 @@
 # a1 and b1 can be missing
 # a2 and b2 can be missing
 
-# this file is now revised to follow the design document hitrun3.Rnw
-# (in the devel directory) rather than the older hitrun.Rnw and hitrun2.Rnw
+# this file is now revised to follow the design document hitrun4.Rnw
+# (in the devel directory) rather than the older hitrun*.Rnw
 
 hitrun <- function(alpha, ...)
     UseMethod("hitrun")
@@ -191,19 +191,21 @@ hitrun.default <- function(alpha, a1 = NULL, b1 = NULL, a2 = NULL, b2 = NULL,
     origin <- foo[is.point, ]
     basis <- foo[is.line, , drop = FALSE]
     basis <- t(basis)
+    # use new orthogonal basis function
+    basis <- qgram(basis)
 
     # at this point
     #     fred <- function(x) origin + basis %*% x
     # maps from new coordinates (NC) onto the affine hull
-    #     of the constraint (a convex polytope) in original coordinated (OC)
+    #     of the constraint (a convex polytope) in original coordinates (OC)
 
     amat <- qneg(hrep4[ , - c(1, 2), drop = FALSE])
     bvec <- hrep4[ , 2]
     bvec <- qmq(bvec, qmatmult(amat, cbind(origin)))
     amat <- qmatmult(amat, basis)
 
     # at this point
-    #     sally <- function(x) all(amat %*% x <= bvec)
+    #     function(x) all(amat %*% x <= bvec)
     # is the indicator function of a convex polytope in NC that is
     # mapped one-to-one onto the constraint set in OC by the function
     # fred defined in the previous comment

inst/doc/pp1.Rnw

@@ -67,7 +67,7 @@ The Polya posterior can be modified to take into account
 available prior information about the population. This leads to 
 the constrained Polya posterior. More discussion about the theory
 underlying the constrained Polya posterior can be found in 
- Lazar, Meeden and Nelson (2005). \nocite{l-m-n05}. The discussion 
+ Lazar, Meeden and Nelson (2008). \nocite{l-m-n05}. The discussion 
 in this vignette will focus on using the constrained Polya 
 posterior to simulate completed copies of the population. 
 
@@ -274,7 +274,7 @@ $\texttt{burnin} - 1$ simulated population means and returns the
 remaining $\texttt{reps} - \texttt{burnin} + 1$
 simulated population means. In this case we are keeping all of the 
 simulated means because $\texttt{burnin} = 1$. 
-We generated a total of  $\texttt{reps}=100,001$ means. The average of these 200,001
+We generated a total of  $\texttt{reps}=200,001$ means. The average of these 200,001
 is given in the output.
 The plot of the first and each successive two hundredth simulated 
 population mean is given 
@@ -317,6 +317,16 @@ out[[2]]
 out[[3]] 
 @
 
+Qu, Meeden and Zhang (2015) \nocite{qmz15}  and Strief and Meeden (2014) \nocite{s-m14}
+discuss two situations where the constrained Polya posterior yields good inference procedures.
+In the first, they demonstrate  that for some small area estimation problems
+the constrained Polya posterior  yields  more robust procedures than standard methods. 
+In the second,  they demonstrate that  for each unit in a sample the constrained Polya 
+posterior can be used to find a weight. This weight can be given the usual interpretation 
+as the number of units in the population the sampled unit represents. These weights 
+can then be used to find good estimates of population parameters.
+
+
 \subsection{Some computing issues}
 
 \subsubsection{How long to run the chain}
@@ -463,13 +473,13 @@ wtpolyap(ysamp,wts,25)
 
 \section{An example using the hitrun function}
 
-A limitation of the function \texttt{constrppprob}  is that it assumes that 
-given the sample the posterior 
-distribution of interest over the polytope defined by the constraints 
-is the uniform distribution. In practice one would like to be able to use 
-an arbitrary Dirichlet distribution as a possible posterior. 
-The \texttt{hitrun} function handles this more general 
-problem.
+
+A limitation of the function \texttt{constrppprob}  is that one cannot add any 
+constraints.  In practice one would like to be able to find approximately
+the expectation of the components of a constrained  Dirichlet distribution.
+The \texttt{hitrun} function handles this more general problem.
+
+
 
 For a survey  sampling example suppose 
 we have a population that is divided into two strata.

tests/hitrun.R

@@ -57,6 +57,7 @@ foo <- vrep3[ , - c(1, 2)]
 origin <- foo[is.point, ]
 basis <- foo[is.line, ]
 basis <- t(basis)
+basis <- qgram(basis)
 identical(q2d(origin), hout$origin)
 identical(q2d(basis), hout$basis)
 

tests/hitrun.Rout.save

@@ -1,7 +1,7 @@
 
-R version 3.1.1 (2014-07-10) -- "Sock it to Me"
-Copyright (C) 2014 The R Foundation for Statistical Computing
-Platform: i686-pc-linux-gnu (32-bit)
+R version 3.4.0 (2017-04-21) -- "You Stupid Darkness"
+Copyright (C) 2017 The R Foundation for Statistical Computing
+Platform: x86_64-pc-linux-gnu (64-bit)
 
 R is free software and comes with ABSOLUTELY NO WARRANTY.
 You are welcome to redistribute it under certain conditions.
@@ -97,6 +97,7 @@ See the Warnings sections added to help pages for
 > origin <- foo[is.point, ]
 > basis <- foo[is.line, ]
 > basis <- t(basis)
+> basis <- qgram(basis)
 > identical(q2d(origin), hout$origin)
 [1] TRUE
 > identical(q2d(basis), hout$basis)
@@ -260,4 +261,4 @@ See the Warnings sections added to help pages for
 > 
 > proc.time()
    user  system elapsed 
-  0.756   0.028   0.790 
+  0.568   0.008   0.573 

vignettes/pp1.Rnw

@@ -67,7 +67,7 @@ The Polya posterior can be modified to take into account
 available prior information about the population. This leads to 
 the constrained Polya posterior. More discussion about the theory
 underlying the constrained Polya posterior can be found in 
- Lazar, Meeden and Nelson (2005). \nocite{l-m-n05}. The discussion 
+ Lazar, Meeden and Nelson (2008). \nocite{l-m-n05}. The discussion 
 in this vignette will focus on using the constrained Polya 
 posterior to simulate completed copies of the population. 
 
@@ -274,7 +274,7 @@ $\texttt{burnin} - 1$ simulated population means and returns the
 remaining $\texttt{reps} - \texttt{burnin} + 1$
 simulated population means. In this case we are keeping all of the 
 simulated means because $\texttt{burnin} = 1$. 
-We generated a total of  $\texttt{reps}=100,001$ means. The average of these 200,001
+We generated a total of  $\texttt{reps}=200,001$ means. The average of these 200,001
 is given in the output.
 The plot of the first and each successive two hundredth simulated 
 population mean is given 
@@ -317,6 +317,16 @@ out[[2]]
 out[[3]] 
 @
 
+Qu, Meeden and Zhang (2015) \nocite{qmz15}  and Strief and Meeden (2014) \nocite{s-m14}
+discuss two situations where the constrained Polya posterior yields good inference procedures.
+In the first, they demonstrate  that for some small area estimation problems
+the constrained Polya posterior  yields  more robust procedures than standard methods. 
+In the second,  they demonstrate that  for each unit in a sample the constrained Polya 
+posterior can be used to find a weight. This weight can be given the usual interpretation 
+as the number of units in the population the sampled unit represents. These weights 
+can then be used to find good estimates of population parameters.
+
+
 \subsection{Some computing issues}
 
 \subsubsection{How long to run the chain}
@@ -463,13 +473,13 @@ wtpolyap(ysamp,wts,25)
 
 \section{An example using the hitrun function}
 
-A limitation of the function \texttt{constrppprob}  is that it assumes that 
-given the sample the posterior 
-distribution of interest over the polytope defined by the constraints 
-is the uniform distribution. In practice one would like to be able to use 
-an arbitrary Dirichlet distribution as a possible posterior. 
-The \texttt{hitrun} function handles this more general 
-problem.
+
+A limitation of the function \texttt{constrppprob}  is that one cannot add any 
+constraints.  In practice one would like to be able to find approximately
+the expectation of the components of a constrained  Dirichlet distribution.
+The \texttt{hitrun} function handles this more general problem.
+
+
 
 For a survey  sampling example suppose 
 we have a population that is divided into two strata.

vignettes/ref.bib

@@ -73,5 +73,27 @@ @article{l-m-n05
 TITLe="A noninformative {B}ayesian approach to finite population
        sampling using auxiliary variables",
 JOURNAL = "Survey Methodology",
+volume = 34,
 PAGES = "51--64",
 year="2008"}
+
+
+@article{qmz15,
+AUTHOR = "Qu, Y. and Meeden, G. and Zhang, B.",
+TITLE = "An objective stepwise Bayes approach to small area estimation",
+JOURNAL = " Journal of Statistical Computation and Simulation",
+PAGES = "1474--1494",
+VOLUME = "85",
+YEAR = "2015"
+}
+
+
+@article{s-m14,
+AUTHOR   = "Strief, J. and Meeden, G.",
+TITLE    =  "Objective Stepwise  Bayes Weights in Survey Sampling",
+JOURNAL  =  "Survey Methodology",
+VOLUME  = "39",
+PAGES  = "1--27",
+YEAR   = "2014"}
+
+