version 2.1

cran · Feb 8, 2018 · fd36a16 · fd36a16
1 parent daccdff
commit fd36a16
Show file tree

Hide file tree

Showing 95 changed files with 2,494 additions and 552 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,16 +1,16 @@
 Package: BayesGOF
 Type: Package
 Title: Bayesian Modeling via Goodness of Fit
-Version: 1.4
-Date: 2018-01-02
+Version: 2.1
+Date: 2018-02-08
 Author: Subhadeep Mukhopadhyay, Douglas Fletcher
 Maintainer: Doug Fletcher <tug25070@temple.edu>
-Description: Non-parametric method for learning prior distribution starting with parametric (subjective) prior.  It performs four interconnected tasks: (i) characterizes the uncertainty of the elicited prior; (ii) exploratory diagnostic for checking prior-data conflict; (iii) computes the final statistical prior density estimate; and (iv) performs macro- and micro-inference. Primary reference is Mukhopadhyay, S. and Fletcher, D. (2017, Technical Report). 
+Description: A Bayesian data modeling scheme that performs four interconnected tasks: (i) characterizes the uncertainty of the elicited parametric prior; (ii) provides exploratory diagnostic for checking prior-data conflict; (iii) computes the final statistical prior density estimate; and (iv) executes macro- and micro-inference. Primary reference is Mukhopadhyay, S. and Fletcher, D. (2018, Technical Report, <arXiv:1802.00474>). 
 Depends: orthopolynom, VGAM
 Suggests: knitr, rmarkdown
 VignetteBuilder: knitr
 License: GPL-2
 NeedsCompilation: no
-Packaged: 2018-01-03 14:43:19 UTC; Doug
+Packaged: 2018-02-08 16:34:52 UTC; Doug
 Repository: CRAN
-Date/Publication: 2018-01-05 19:00:44 UTC
+Date/Publication: 2018-02-08 17:23:47 UTC
diff --git a/MD5 b/MD5
@@ -1,55 +1,108 @@
-ff715b03553ac6270dbb83c08e7f16df *DESCRIPTION
-a4f3d575676f1b80e1058c92810fd2b7 *NAMESPACE
-e8cdb771de69edc672a7aac8d9440bb9 *R/BetaBinoMLE.R
-fe43131dc70d9bfb435a4e0dcd3ae58d *R/ConMean.prt.2.R
-f534a2cee775d2300a120951ff633f4a *R/DS.PostMean.R
-6f77a33e7e6f0fa1184b966f73dae483 *R/DS.PostMean.reduce.R
-df69695935b3e562e17b49c8223033a9 *R/DS.macro.inf.R
-d7e1d4d09924ca058faacdad2455fa6d *R/DS.micro.inf.R
-60be8417bea981f785e722ec25c278b9 *R/DS.mode.map.R
-03be355b3798c6a69b97548e5eee3926 *R/DS.mode.reduce.R
-8f5455dc47b5866c8c6423ee6b5f83ab *R/DS.prior.R
+8c77c3c19d95daffd0bf9cce56bff017 *DESCRIPTION
+388fd6d9b2d5c4ac4f8c83332bd163ba *NAMESPACE
+e8874fa248257c1fde8bcc55c8c64103 *R/ConMean.prt.2.bb.R
+dd9822d2b54974880c9801edc44b662f *R/ConMean.prt.2.nn.R
+3eb68f5507e9d541427effca082ff250 *R/ConMean.prt.2.pg.R
+f210a56b1292fb4ddaaed0cebe0d8e2a *R/DS.PostMean.R
+26410766099872ebc468aa5bd229a0eb *R/DS.PostMean.bbu.R
+e425b9419c3889396aa84b80125fe188 *R/DS.PostMean.nnu.R
+94c22af206dd01d07486ed9edd463d43 *R/DS.PostMean.pge.R
+7c0476b19773fdb2396929384a31ded3 *R/DS.PostMean.pgu.R
+ed4741f3f4462ab382c358f42a5fcdeb *R/DS.PostMean.reduce.R
+177dc6c20a6d45fd3aaaa2d95fe6fb24 *R/DS.PostMean.reduce.bbu.R
+11cc03c067f4ab80bc1962b8bf485e73 *R/DS.PostMean.reduce.nnu.R
+4e294de31b7f0428d6e512b1dec4a774 *R/DS.PostMean.reduce.pgu.R
+35ce329a6610dd7a4383a9e9926556e3 *R/DS.PostMode.R
+5e05ecb42d278459380074c28a111d6c *R/DS.PostMode.reduce.R
+642e34bded52c4dc9533784e965491f8 *R/DS.macro.inf.R
+c00328ab3fb487244b8db34262a7476a *R/DS.macro.inf.bbu.R
+4ddabf948f321d35aeea31963be13d80 *R/DS.macro.inf.nnu.R
+f2d0cd7ad4e53783ac25c1fb701ff374 *R/DS.macro.inf.pge.R
+d7993cbcdadf2772361f3fefa6f60cdb *R/DS.macro.inf.pgu.R
+131c10e579f32e6a389065b197fb24b1 *R/DS.micro.inf.R
+748eefb997234943c67f195b94ddde43 *R/DS.micro.inf.bbu.R
+f4c077d4bd0724119a1d209023ccca6c *R/DS.micro.inf.nnu.R
+0f58226b46c55f3dcae5e6d0f46daa97 *R/DS.micro.inf.pge.R
+b2cbba64c9b9d39ee7e75e4227b3bfd5 *R/DS.micro.inf.pgu.R
+b9408cd72a0a4304f50a7aeb87ba0be4 *R/DS.mode.map.bbu.R
+fb3141620abc5147dcb0dd42e9cd12d7 *R/DS.mode.map.nnu.R
+34c4cb9aebfdc34b6e4b01a570f7ee2c *R/DS.mode.map.pgu.R
+3fb4119ff1202e4db7ba525244522a55 *R/DS.mode.reduce.bbu.R
+a03ddf0bf44d6dadad746616c3e2c7ca *R/DS.mode.reduce.nnu.R
+45588e46af107453e3098d2f6d54619e *R/DS.mode.reduce.pgu.R
+a45d1342794901bb53f82d3ef23a3486 *R/DS.prior.R
+276cde93a9e1361ac97e61d2dd67cf8c *R/DS.prior.bbu.R
+792fbc5b79305a01ccef789f3725f6d2 *R/DS.prior.nnu.R
+840689ac8dc209ea2b686b46b1d55c05 *R/DS.prior.pgu.R
+59761b704e4c2594809a9bc6bdb6ab0d *R/DS.sampler.R
+79c5ea8e557814c77ab4f8114ba3e650 *R/DS.sampler.post.R
 e5354a65b65d2c5b81582559ee7a1030 *R/EXP.denom.R
 5ded1e76591fbd05471de371c5261042 *R/EXP.num.1.R
 cd2cdbd50f105ec57eb2fd32efc9630d *R/EXP.num.2.R
-399a1ec91a0e02f8186b74648054c169 *R/EXP.score.R
-e04999cc772d27e577942b140b77ef24 *R/LP.basis.beta.R
+9bb3353c77f1bcb5c1e5dc9c1780c1a3 *R/EXP.score.R
+e1b8297394b2aeb30dac851894205ece *R/LP.basis.beta.R
+b8b1dbd2309ba601d7188a889c96b9ca *R/LP.basis.gamma.R
+ec871cf587588dc9330a3f9d1420f9f8 *R/LP.basis.norm.R
 0488b1290b901beda5de0d54c6000679 *R/LP.smooth.R
 6b2a90b2028ddec70ff0a68fa1a519e0 *R/Local.Mode.R
-cc0c13114ef93fab90e4634341fa76ce *R/Reduce.LP.coef.R
+51d4b6e256174f370f2e8c7e3586c693 *R/Reduce.LP.coef.univ.R
+0d6ac55da028d87e14a5035e36d16232 *R/gLP.basis.R
+d4a2a8031c684b6364d3d6f442a7829e *R/gMLE.bb.R
+43e47e540af3ec33f977cd5475c20953 *R/gMLE.nn.R
+06c0cb40197d168f55c0b86a50c92bdc *R/gMLE.pg.R
+4575aa9706b6a5fd8ed35464a098852c *R/lambda.i.R
 7fa03fde303c925b5de4e3fa82bfe8a4 *R/mom.est.R
-71c42d22a062530811bf5774eab9b56f *R/plot.DS_GF.R
-1f8be9e42445840b66f7f2457fa95f15 *R/plot.DS_GF_macro_mean.R
-22ffc8f09ab875efedf86b9f35707f27 *R/plot.DS_GF_macro_mode.R
-ca68dbc678d9f573c109a904b2ac29fb *R/plot.DS_GF_micro.R
+aa58cc42ac6447acbd5611010a75dc02 *R/plot.DS_GF.R
+c7d0dfb2d0f4407384c763d664ac2584 *R/plot.DS_GF_macro_mean.R
+dbf1a306ed4278f8e312298804a1ff12 *R/plot.DS_GF_macro_mode.R
+4657fd0fb50bdca21802f2e910b5be47 *R/plot.DS_GF_micro.R
 32731dab82fe5d939ab07f068eece6e5 *R/print.DS_GF.R
 847051f9d55ea4d3df293cbe661e8fb7 *R/print.DS_GF_macro_mean.R
 5d2acbda0d40b6b8ea5074459f8916ba *R/print.DS_GF_macro_mode.R
 9a2ea0cbaae526bf289d047e08c484c2 *R/print.DS_GF_micro.R
-c5751d11df67d3e13765bb79a0ed8cdf *R/rDS.R
-904d66c45aae0007a9a9790007834619 *R/rDS.posterior.R
-43346db63edf02693e6dc6cde4561463 *R/weight.fun.beta.R
-52e60785b26ceed34594689813d7cde2 *build/partial.rdb
+e41c1edfd9b441d09fe161602ec14b6b *R/rDS.bbu.R
+1187f5e7ee3a75125383a9f35c4f3d46 *R/rDS.nnu.R
+9f50dd2ce26a87e2d93c8538ade2a216 *R/rDS.pgu.R
+e50e2e8cbcc87458078edb6758f4772c *R/rDS.post.bbu.R
+e756da3965695fb702859d9ab16c8948 *R/rDS.post.nnu.R
+cda2b544f890d24da352497f088fedb3 *R/rDS.post.pgu.R
+e8dd3d9cdc01501517963ff94c6fa5b9 *R/weight.fun.univ.R
+c6ebf4aa378cdb5bbbb6ab4d7f5a6421 *build/partial.rdb
 32bb6c15cc261cc5a800feaef03df159 *build/vignette.rds
+2235da02896248caee37a6bfd8f1675e *data/AutoIns.rda
+890414d6043ab45f0fc2d487c774eb93 *data/ChildIll.rda
+5597ec356f1189688b399a8835e03316 *data/CorbBfly.rda
+b4ec2d57f4f400e72d9bf5026e030d42 *data/NorbergIns.rda
+f66e2bfb6773231c37acda708ae6f320 *data/arsenic.rda
 4eb418b2906c27f4d7c4f1820cdfef39 *data/rat.rda
 86f2e44eba816a17d1c21b98238eb0f9 *data/ship.rda
 2ea567712e420d7dad930065d164f0ea *data/surg.rda
 b2cc985b167635ee9460b40031cb97b1 *data/tacks.rda
 d5e17a0af9b0f694d178a824213f7f04 *data/terb.rda
-b6b93be51774b48490af693f8da7b977 *inst/doc/vignette_BayesGOF.R
-cbd58b4bd918f85fd591770a32884297 *inst/doc/vignette_BayesGOF.Rmd
-e4d0976b6350f76127b0c4229a87fbc2 *inst/doc/vignette_BayesGOF.html
-f602b20ce666636fd4ce1fe3e2e4b14f *man/BayesGOF-package.Rd
-d5fe161c86340df185b1bf8722ed3244 *man/BetaBinoMLE.Rd
-b14a54fa25faf8496826ef5e2a0938aa *man/DS.PostMean.Rd
-f212f760e4377b14adde601a86a6c06b *man/DS.macro.inf.Rd
-323c51a0c260a624791b4c2a5cd96081 *man/DS.micro.inf.Rd
-883a6c582feafd1a0c212e75a234c579 *man/DS.prior.Rd
-80f2637e9f992da56001c6e04a910c57 *man/LP.basis.beta.Rd
-c5ae2041c3448673d570d4f62292a08a *man/rDS.Rd
-d0e4b41b002567b8dd89ece7943ad973 *man/rat.Rd
-036e74bea69c9cb18ce9435abda7a77d *man/ship.Rd
+cc1812eb703a57cffe97326febc8574e *data/ulcer.rda
+aae646e1ab0b14b61ce2c13c3e9798b9 *inst/doc/vignette_BayesGOF.R
+00477fab27864a039894e86b46e073b8 *inst/doc/vignette_BayesGOF.Rmd
+fa7e4dd971125091a2fb354d13fdcb9d *inst/doc/vignette_BayesGOF.html
+a41b5099796db2c2d56fe2f6db433a2c *man/AutoIns.Rd
+4568ed5c592b8b291f9b7f667c19f786 *man/BayesGOF-package.Rd
+e1aa854297548a3d819cf1107e552745 *man/ChildIll.Rd
+69425426d612b0624b71ba777c13ce8a *man/CorbBfly.Rd
+d225ad0e6c5fa07a33d7de5e3c88af24 *man/DS.PostMean.Rd
+173b507c0652bdefb19c078df65ba3a7 *man/DS.PostMode.Rd
+c62d319e6dd8a42576930e2d7b8c56c2 *man/DS.macro.inf.Rd
+a4fe6f3b4a2a9b557d737718a459a232 *man/DS.micro.inf.Rd
+1dec0080725d7f66422d830c83c287a1 *man/DS.prior.Rd
+8f2b93605ca690f4dd8fcba1e84cdb3e *man/DS.sampler.Rd
+1ab0aed54cbe4c0ca2c93ff114f4a254 *man/NorbergIns.Rd
+381f94e2b04e233cef79447defd45cbc *man/arsenic.Rd
+3bb7991e0dfb8defa05a1ff8c2bcc556 *man/gLP.basis.Rd
+a9fe0929f4db5dc101a80867382678e2 *man/gMLE.bb.Rd
+e829cd658710bb200b205aa6cd258a21 *man/gMLE.nn.Rd
+c5dfe04006698e33388e22b6a83dc1a0 *man/gMLE.pg.Rd
+acbf247e0e7ceb171e4b9d88059ddca3 *man/rat.Rd
+5df7768440e333efe4171b7830520775 *man/ship.Rd
 1a4578d3061e06f366a3a81187e2d7bb *man/surg.Rd
-7df318a497e78c94fafa22de2db27802 *man/tacks.Rd
+9081a09d6db42cfda219ec187d66122d *man/tacks.Rd
 e260deea72882c143a52313389aa8b6d *man/terb.Rd
-cbd58b4bd918f85fd591770a32884297 *vignettes/vignette_BayesGOF.Rmd
+b2b1af83da55655cd17d0886f6090b71 *man/ulcer.Rd
+00477fab27864a039894e86b46e073b8 *vignettes/vignette_BayesGOF.Rmd
diff --git a/NAMESPACE b/NAMESPACE
@@ -1,7 +1,8 @@
 exportPattern("^[[:alpha:]]+")
 import(VGAM, orthopolynom)
 importFrom("stats", "dbeta", "optim", "pbeta", "qbeta", "rbeta",
-             "rbinom", "runif", "sd", "var")
+             "rbinom", "runif", "sd", "var","dgamma", "dnorm",  
+			 "pgamma", "pnorm", "qgamma", "qnorm", "rgamma", "rnorm", "rpois")
 importFrom("graphics", "abline", "axis", "legend", "lines", "par",
              "points", "title")
 importFrom("utils", "installed.packages")

diff --git a/R/ConMean.prt.2.R → R/ConMean.prt.2.bb.R b/R/ConMean.prt.2.R → R/ConMean.prt.2.bb.R
@@ -1,8 +1,8 @@
-ConMean.prt.2 <-
+ConMean.prt.2.bb <-
 function(c.vec, weight, leg.mat, u, par1, par2){
-	#### theta is G inv u = qnorm(u,muhat,tauhat)
+	#### theta is G inv u = qbeta(u,par1, par2)
 	inner.part <- qbeta(u, par1, par2)*leg.mat
 	v <-(t(inner.part)%*% weight)/nrow(leg.mat)
 	num.2 <- as.vector(v) %*% c.vec
 	return(num.2)
-}
+}
diff --git a/R/ConMean.prt.2.nn.R b/R/ConMean.prt.2.nn.R
@@ -0,0 +1,8 @@
+ConMean.prt.2.nn <-
+function(c.vec, weight, leg.mat, u, par1, par2){
+	#### theta is G inv u = qnorm(u,muhat,tauhat)
+	inner.part <- qnorm(u, par1, sd=sqrt(par2))*leg.mat
+	v <-(t(inner.part)%*% weight)/nrow(leg.mat)
+	num.2 <- as.vector(v) %*% c.vec
+	return(num.2)
+}
diff --git a/R/ConMean.prt.2.pg.R b/R/ConMean.prt.2.pg.R
@@ -0,0 +1,8 @@
+ConMean.prt.2.pg <- 
+function(c.vec, weight, leg.mat, u, par1, par2){
+	#### theta is G inv u = qgamma(u,muhat,tauhat)
+	inner.part <- qgamma(u, par1, scale = par2)*leg.mat
+	v <-(t(inner.part)%*% weight)/nrow(leg.mat)
+	num.2 <- as.vector(v) %*% c.vec
+	return(num.2)
+}
diff --git a/R/DS.PostMean.R b/R/DS.PostMean.R
@@ -1,14 +1,21 @@
 DS.PostMean <-
-function(y.i, n.i, g.par, u, LP.par){
-		post.par1.i <- y.i + g.par[1]
-		post.par2.i <- n.i - y.i + g.par[2]
-		weight <- weight.fun.beta(u, g.par[1], g.par[2], post.par1.i, post.par2.i)
-		leg.mat <- LP.basis.beta(u,c(1,1), length(LP.par))
-		if(sum(LP.par^2) == 0){
-				ConMean.ind <- (post.par1.i)/(post.par1.i + post.par2.i)
+function(y.0, n.0 = NULL, g.par, LP.par, B = 250, 
+		 family = c("Normal","Binomial", "Poisson"), e.0 = NULL){
+	fam = match.arg(family)
+	u <- seq(1/B, 1-1/B, length.out = B)
+	switch(fam,
+		"Normal" = {
+			DS.PostMean.nnu(y.i = y.0, se.i=n.0, g.par = g.par, u=u,LP.par = LP.par)
+		 },
+		 "Binomial" = {
+			DS.PostMean.bbu(y.i = y.0, n.i=n.0,g.par = g.par, u=u,LP.par = LP.par)
+		 },
+		 "Poisson" = {
+			if(is.null(e.0) == TRUE){
+				DS.PostMean.pgu(x.i = y.0, g.par = g.par , u = u, LP.par = LP.par)
 				} else {
-				post.mu <- post.par1.i /(post.par1.i + post.par2.i)
-				ConMean.ind <- ( post.mu + ConMean.prt.2(LP.par, weight, leg.mat, u, g.par[1], g.par[2]) ) / EXP.denom(LP.par,weight,leg.mat)
+				DS.PostMean.pge(x.i = y.0, e.i = e.0, g.par = g.par, u = u, LP.par = LP.par)
 				}
-		return(ConMean.ind)
-	}
+				}
+		)
+	}
diff --git a/R/DS.PostMean.bbu.R b/R/DS.PostMean.bbu.R
@@ -0,0 +1,15 @@
+DS.PostMean.bbu <-
+function(y.i, n.i, g.par, u, LP.par){
+		fam <- "Binomial"
+		post.par1.i <- y.i + g.par[1]
+		post.par2.i <- n.i - y.i + g.par[2]
+		weight <- weight.fun.univ(u, g.par[1], g.par[2], post.par1.i, post.par2.i, family = fam)
+		leg.mat <- LP.basis.beta(u,c(1,1), length(LP.par))
+			if(sum(LP.par^2) == 0){
+				ConMean.ind <- (post.par1.i)/(post.par1.i + post.par2.i)
+				} else {
+				post.mu <- post.par1.i /(post.par1.i + post.par2.i)
+				ConMean.ind <- ( post.mu + ConMean.prt.2.bb(LP.par, weight, leg.mat, u, g.par[1], g.par[2]) ) / EXP.denom(LP.par,weight,leg.mat)
+				}
+			return(ConMean.ind)
+		}
diff --git a/R/DS.PostMean.nnu.R b/R/DS.PostMean.nnu.R
@@ -0,0 +1,15 @@
+DS.PostMean.nnu <-
+function(y.i, se.i, g.par, u, LP.par){
+		fam <- "Normal"
+		post.mu.i <- lambda.i(se.i, g.par[2]) * g.par[1] +
+					(1-lambda.i(se.i, g.par[2]))* y.i
+		post.tau2.i <- (1-lambda.i(se.i, g.par[2]))*se.i^2 #output is VARIANCE
+		weight <- weight.fun.univ(u, g.par[1], g.par[2], post.mu.i, post.tau2.i, family = fam)
+		leg.mat <- LP.basis.beta(u,c(1,1), length(LP.par))
+		if(sum(LP.par^2) == 0){
+				ConMean.ind <- post.mu.i
+				} else {
+				ConMean.ind <- ( post.mu.i + ConMean.prt.2.nn(LP.par, weight, leg.mat, u, g.par[1], g.par[2]) ) / EXP.denom(LP.par,weight,leg.mat)
+				}
+			return(ConMean.ind)
+	}
diff --git a/R/DS.PostMean.pge.R b/R/DS.PostMean.pge.R
@@ -0,0 +1,13 @@
+DS.PostMean.pge <- 
+function(x.i, e.i, g.par, u, LP.par){
+	fam <- "Poisson"
+	B <- 250
+	post.par1.i <- x.i + g.par[1]
+	post.par2.i <- g.par[2]/(1+e.i*g.par[2])
+	#u <- seq(1/B,1-(1/B), length.out = B) 
+	weight <- weight.fun.univ(u, g.par[1],g.par[2],post.par1.i,post.par2.i, family = fam)
+	leg.mat <- LP.basis.beta(u, c(1,1), length(LP.par))
+	post.mu <- post.par1.i * post.par2.i
+	ConMean.ind <- (post.mu + ConMean.prt.2.pg(LP.par, weight, leg.mat, u, g.par[1], g.par[2]) ) / EXP.denom(LP.par,weight,leg.mat)
+	return(ConMean.ind)	
+}
diff --git a/R/DS.PostMean.pgu.R b/R/DS.PostMean.pgu.R
@@ -0,0 +1,15 @@
+DS.PostMean.pgu <-
+function(x.i, g.par, u, LP.par){
+		fam <- "Poisson"
+		post.par1.i <- x.i + g.par[1]
+		post.par2.i <- g.par[2]/(1+g.par[2])
+		weight <- weight.fun.univ(u, g.par[1],g.par[2],post.par1.i,post.par2.i, family = fam)
+		leg.mat <- LP.basis.beta(u, c(1,1), length(LP.par))
+		if(sum(LP.par^2) == 0){
+				ConMean.ind <- post.par1.i*post.par2.i
+				} else {
+				post.mu <- post.par1.i*post.par2.i
+				ConMean.ind <- (post.mu + ConMean.prt.2.pg(LP.par, weight, leg.mat, u, g.par[1], g.par[2]) ) / EXP.denom(LP.par,weight,leg.mat)
+				}
+			return(ConMean.ind)
+		}
diff --git a/R/DS.PostMean.reduce.R b/R/DS.PostMean.reduce.R
@@ -1,19 +1,15 @@
 DS.PostMean.reduce <-
 function(DS.GF.obj){
-#####################################################
-# INPUTS
-#  DS.GF.obj		AutoBayes-DataCorrect object
-# OUTPUTS
-#  ConMean.vec		vector of conditional mean of entire sample
-	y.i <- DS.GF.obj$obs.data[,1]
-	n.i <- DS.GF.obj$obs.data[,2]
-	if(sum(DS.GF.obj$LP.par^2) == 0){
-		ConMean.vec <- (DS.GF.obj$g.par[1]+y.i)/(DS.GF.obj$g.par[1]+DS.GF.obj$g.par[2]+n.i)
-		} else {
-		B <- dim(DS.GF.obj$prior.data)[1]
-		u <- seq(1/B,1-(1/B), length.out = B) ##unit interval, 0 to 1
-		ConMean.vec <- NULL
-		ConMean.vec <- apply(DS.GF.obj$obs.data, 1, function(x) DS.PostMean(x[1], x[2], DS.GF.obj$g.par, u, DS.GF.obj$LP.par))
+	fam = DS.GF.obj$fam
+	switch(fam,
+		"Normal" = {
+			DS.PostMean.reduce.nnu(DS.GF.obj)
+		 },
+		 "Binomial" = {
+			DS.PostMean.reduce.bbu(DS.GF.obj)
+		 },
+		 "Poisson" = {
+			DS.PostMean.reduce.pgu(DS.GF.obj)				
+			 }
+		)
 	}
-	return(ConMean.vec)
-}
diff --git a/R/DS.PostMean.reduce.bbu.R b/R/DS.PostMean.reduce.bbu.R
@@ -0,0 +1,19 @@
+DS.PostMean.reduce.bbu <-
+function(DS.GF.obj){
+#####################################################
+# INPUTS
+#  DS.GF.obj		AutoBayes-DataCorrect object
+# OUTPUTS
+#  ConMean.vec		vector of conditional mean of entire sample
+	y.i <- DS.GF.obj$obs.data[,1]
+	n.i <- DS.GF.obj$obs.data[,2]
+	if(sum(DS.GF.obj$LP.par^2) == 0){
+		ConMean.vec <- (DS.GF.obj$g.par[1]+y.i)/(DS.GF.obj$g.par[1]+DS.GF.obj$g.par[2]+n.i)
+		} else {
+		B <- dim(DS.GF.obj$prior.fit)[1]
+		u <- seq(1/B,1-(1/B), length.out = B) ##unit interval, 0 to 1
+		ConMean.vec <- NULL
+		ConMean.vec <- apply(DS.GF.obj$obs.data, 1, function(x) DS.PostMean.bbu(x[1], x[2], DS.GF.obj$g.par, u, DS.GF.obj$LP.par))
+	}
+	return(ConMean.vec)
+}
diff --git a/R/DS.PostMean.reduce.nnu.R b/R/DS.PostMean.reduce.nnu.R
@@ -0,0 +1,20 @@
+DS.PostMean.reduce.nnu <-
+function(DS.GF.obj){
+#####################################################
+# INPUTS
+#  DS.GF.obj		AutoBayes-DataCorrect object
+# OUTPUTS
+#  ConMean.vec		vector of conditional mean of entire sample
+	y.vec <- DS.GF.obj$obs.data[,1]
+	se.vec <- DS.GF.obj$obs.data[,2]
+	if(sum(DS.GF.obj$LP.par^2) == 0){
+		ConMean.vec <- lambda.i(se.vec, DS.GF.obj$g.par[2]) * DS.GF.obj$g.par[1] +
+					(1-lambda.i(se.vec, DS.GF.obj$g.par[2]))* y.vec
+		} else {
+		B <- dim(DS.GF.obj$prior.fit)[1]
+		u <- seq(1/B,1-(1/B), length.out = B) ##unit interval, 0 to 1
+		ConMean.vec <- NULL
+		ConMean.vec <- apply(DS.GF.obj$obs.data, 1, function(x) DS.PostMean.nnu(x[1], x[2], DS.GF.obj$g.par, u, DS.GF.obj$LP.par))
+	}
+	return(ConMean.vec)
+}
diff --git a/R/DS.PostMean.reduce.pgu.R b/R/DS.PostMean.reduce.pgu.R
@@ -0,0 +1,19 @@
+DS.PostMean.reduce.pgu <-
+function(DS.GF.obj){
+#####################################################
+# INPUTS
+#  DS.GF.obj		AutoBayes-DataCorrect object
+# OUTPUTS
+#  ConMean.vec		vector of conditional mean of entire sample
+	tbl <- table(DS.GF.obj$obs.data)
+	cnt.vec <- as.integer(unlist(dimnames(tbl)))
+	if(sum(DS.GF.obj$LP.par^2) == 0){
+		ConMean.vec <- (DS.GF.obj$g.par[1]+cnt.vec)*(DS.GF.obj$g.par[2])/(1+DS.GF.obj$g.par[2])
+		} else {
+		B <- dim(DS.GF.obj$prior.fit)[1]
+		u <- seq(1/B,1-(1/B), length.out = B) ##unit interval, 0 to 1
+		ConMean.vec <- NULL
+		ConMean.vec <- sapply(cnt.vec, function(x) DS.PostMean.pgu(x, DS.GF.obj$g.par, u, DS.GF.obj$LP.par))
+	}
+	return(ConMean.vec)
+}