diff --git a/analysis/DESCRIPTION b/analysis/DESCRIPTION
new file mode 100644
index 00000000..b438cad5
--- /dev/null
+++ b/analysis/DESCRIPTION
@@ -0,0 +1,10 @@
+Package: rappor
+Title: R package for Rappor Analysis
+Description: R functions used for analyzing Rappor data.
+Author: Rappor Discuss <rappor-discuss@googlegroups.com>
+Maintainer: Rappor Discuss <rappor-discuss@googlegroups.com>
+Copyright: Copyright (C) 2014 Google, Inc.
+Version: 0.1
+License: Apache License 2.0
+Url: https://github.com/google/rappor
+Depends: Matrix, RUnit, glmnet, limSolve, parallel
diff --git a/analysis/NAMESPACE b/analysis/NAMESPACE
new file mode 100644
index 00000000..6635d4c6
--- /dev/null
+++ b/analysis/NAMESPACE
@@ -0,0 +1,6 @@
+import(Matrix)
+import(RUnit)
+import(glmnet)
+import(limSolve)
+import(parallel)
+exportPattern("^[^\\.]")
diff --git a/analysis/R/alternative.R b/analysis/R/alternative.R
index 3f0e66d3..8cb6442f 100755
--- a/analysis/R/alternative.R
+++ b/analysis/R/alternative.R
@@ -12,9 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-library(limSolve)
-library(Matrix)
-
 # The next two functions create a matrix (G) and a vector (H) encoding
 # linear inequality constraints that a solution vector (x) must satisfy:
 #                       G * x >= H
@@ -80,4 +77,4 @@ ConstrainedLinModel <- function(X,Y) {
   names(coefs) <- colnames(X)
 
   coefs
-}
\ No newline at end of file
+}
diff --git a/analysis/R/association.R b/analysis/R/association.R
index d1c7b5ee..7b2219cc 100755
--- a/analysis/R/association.R
+++ b/analysis/R/association.R
@@ -12,16 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-library(parallel)  # mclapply
-
-source.rappor <- function(rel_path)  {
-  abs_path <- paste0(Sys.getenv("RAPPOR_REPO", ""), rel_path)
-  source(abs_path)
-}
-
-source.rappor("analysis/R/util.R")  # for Log
-source.rappor("analysis/R/decode.R")  # for ComputeCounts
-
 #
 # Tools used to estimate variable distributions of up to three variables
 #     in RAPPOR. This contains the functions relevant to estimating joint
diff --git a/analysis/R/decode.R b/analysis/R/decode.R
index 7d83f2b9..4ee2a223 100755
--- a/analysis/R/decode.R
+++ b/analysis/R/decode.R
@@ -12,18 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-#
-# This library implements the RAPPOR marginal decoding algorithms using LASSO.
-
-library(glmnet)
-
-# So we don't have to change pwd
-source.rappor <- function(rel_path)  {
-  abs_path <- paste0(Sys.getenv("RAPPOR_REPO", ""), rel_path)
-  source(abs_path)
-}
-
-source.rappor('analysis/R/alternative.R')
+# Implements the RAPPOR marginal decoding algorithms using LASSO.
 
 EstimateBloomCounts <- function(params, obs_counts) {
   # Estimates the number of times each bit in each cohort was set in original
diff --git a/analysis/R/decode_ngrams.R b/analysis/R/decode_ngrams.R
index e2585cbd..bf67f5fa 100755
--- a/analysis/R/decode_ngrams.R
+++ b/analysis/R/decode_ngrams.R
@@ -49,7 +49,6 @@ FindPairwiseCandidates <- function(report_data, N, ngram_params, params) {
 
   .ComputeDist <- function(i, inds, cohorts, reports, maps, params,
                            num_ngrams_collected) {
-    library(glmnet)
     ind <- inds[[i]]
     cohort_subset <- lapply(1:num_ngrams_collected, function(x)
                             cohorts[ind])
diff --git a/analysis/R/ngrams_simulation.R b/analysis/R/ngrams_simulation.R
index ca7ce49b..2bcd39b0 100755
--- a/analysis/R/ngrams_simulation.R
+++ b/analysis/R/ngrams_simulation.R
@@ -18,15 +18,6 @@
 #     dictionary of terms over which we want to learn a distribution. This
 #     mostly contains functions that aid in the generation of synthetic data.
 
-library(RUnit)
-library(parallel)
-
-source("analysis/R/encode.R")
-source("analysis/R/decode.R")
-source("analysis/R/simulation.R")
-source("analysis/R/association.R")
-source("analysis/R/decode_ngrams.R")
-
 # The alphabet is the set of all possible characters that will appear in a
 #     string. Here we use the English alphabet, but one might want to include
 #     numbers or punctuation marks.
@@ -240,11 +231,6 @@ EstimateDictionaryTrial <- function(N, str_len, num_strs,
   #
   # Returns:
   #   List with recovered and true marginals.
-
-  # We call the needed libraries here in order to make them available when this
-  #     function gets called by BorgApply. Otherwise, they do not get included.
-  library(glmnet)
-  library(parallel)
   sim <- SimulateNGrams(N, ngram_params, str_len, num_strs = num_strs,
                         alphabet, params, distribution)
 
diff --git a/analysis/R/read_input.R b/analysis/R/read_input.R
index 47f8be5d..5f79d686 100755
--- a/analysis/R/read_input.R
+++ b/analysis/R/read_input.R
@@ -15,16 +15,6 @@
 #
 # Read parameter, counts and map files.
 
-library(Matrix)
-
-source.rappor <- function(rel_path)  {
-  abs_path <- paste0(Sys.getenv("RAPPOR_REPO", ""), rel_path)
-  source(abs_path)
-}
-
-source.rappor("analysis/R/util.R")  # for Log
-
-
 ReadParameterFile <- function(params_file) {
   # Read parameter file. Format:
   # k, h, m, p, q, f
diff --git a/analysis/R/simulation.R b/analysis/R/simulation.R
index 251c5959..27b194dc 100755
--- a/analysis/R/simulation.R
+++ b/analysis/R/simulation.R
@@ -16,9 +16,6 @@
 # RAPPOR simulation library. Contains code for encoding simulated data and
 #     creating the map used to encode and decode reports.
 
-library(glmnet)
-library(parallel)  # mclapply
-
 SetOfStrings <- function(num_strings = 100) {
   # Generates a set of strings for simulation purposes.
   strs <- paste0("V_", as.character(1:num_strings))
diff --git a/bin/decode_assoc.R b/bin/decode_assoc.R
index 58e35f25..f7993a0c 100755
--- a/bin/decode_assoc.R
+++ b/bin/decode_assoc.R
@@ -113,26 +113,12 @@ if (!interactive()) {
   opts <- ParseOptions()
 }
 
-#
-# Load libraries and source our own code.
-#
-
-library(RJSONIO)  # toJSON()
-
-# So we don't have to change pwd
-source.rappor <- function(rel_path)  {
-  abs_path <- paste0(Sys.getenv("RAPPOR_REPO", ""), rel_path)
-  source(abs_path)
-}
-
-source.rappor("analysis/R/association.R")
-source.rappor("analysis/R/fast_em.R")
-source.rappor("analysis/R/read_input.R")
-source.rappor("analysis/R/util.R")
-
 options(stringsAsFactors = FALSE)
 options(max.print = 100)  # So our structure() debug calls look better
 
+library(RJSONIO)  # toJSON()
+library(rappor)
+
 CreateAssocStringMap <- function(all_cohorts_map, params) {
   # Processes the maps loaded using ReadMapFile and turns it into something
   # that association.R can use.  Namely, we want a map per cohort.
diff --git a/bin/decode_dist.R b/bin/decode_dist.R
index 5c83f741..05e219e2 100755
--- a/bin/decode_dist.R
+++ b/bin/decode_dist.R
@@ -56,25 +56,10 @@ if (!interactive()) {
   opts <- ParseOptions()
 }
 
-#
-# Load libraries and source our own code.
-#
+options(stringsAsFactors = FALSE)
 
 library(RJSONIO)
-
-# So we don't have to change pwd
-source.rappor <- function(rel_path)  {
-  abs_path <- paste0(Sys.getenv("RAPPOR_REPO", ""), rel_path)
-  source(abs_path)
-}
-
-source.rappor("analysis/R/read_input.R")
-source.rappor("analysis/R/decode.R")
-source.rappor("analysis/R/util.R")
-
-source.rappor("analysis/R/alternative.R")
-
-options(stringsAsFactors = FALSE)
+library(rappor)
 
 
 main <- function(opts) {
diff --git a/build.sh b/build.sh
index 44c5cd51..740bd634 100755
--- a/build.sh
+++ b/build.sh
@@ -111,10 +111,16 @@ cpp-client() {
   popd
 }
 
+r-package() {
+  R CMD BUILD analysis/
+  sudo R CMD INSTALL rappor_0.1.tar.gz
+}
+
 if test $# -eq 0 ; then
   cpp-client
   doc
   fastrand
+  r-package
 else
   "$@"
 fi
diff --git a/analysis/R/association_test.R b/tests/RUnit/association_test.R
similarity index 98%
rename from analysis/R/association_test.R
rename to tests/RUnit/association_test.R
index 0cd24ce0..85abf081 100755
--- a/analysis/R/association_test.R
+++ b/tests/RUnit/association_test.R
@@ -15,12 +15,7 @@
 # Authors: vpihur@google.com (Vasyl Pihur), fanti@google.com (Giulia Fanti)
 
 library(RUnit)
-source("analysis/R/encode.R")
-source("analysis/R/decode.R")
-source("analysis/R/simulation.R")
-source("analysis/R/association.R")
-source("analysis/R/fast_em.R")
-source("analysis/R/util.R")
+library(rappor)
 
 SamplePopulations <- function(N, num_variables = 1, params,
                               variable_opts) {
diff --git a/analysis/R/decode_test.R b/tests/RUnit/decode_test.R
similarity index 99%
rename from analysis/R/decode_test.R
rename to tests/RUnit/decode_test.R
index 74c46ce4..83ef4373 100755
--- a/analysis/R/decode_test.R
+++ b/tests/RUnit/decode_test.R
@@ -15,10 +15,10 @@
 
 library(RUnit)
 library(abind)
-
-source('analysis/R/decode.R')
+library(rappor)
 source('tests/gen_counts.R')
 
+
 L1Distance <- function(X, Y) {
   # Computes the L1 distance between two named vectors
   common <- intersect(names(X), names(Y))
diff --git a/analysis/R/run_tests.R b/tests/RUnit/run_tests.R
similarity index 95%
rename from analysis/R/run_tests.R
rename to tests/RUnit/run_tests.R
index 8a4692fa..9ffb50be 100755
--- a/analysis/R/run_tests.R
+++ b/tests/RUnit/run_tests.R
@@ -17,10 +17,8 @@
 #
 # Run unit tests for RAPPOR R code.
 
-library(RUnit)
-
 run_tests <- function() {
-  dirs <- "analysis/R"  # Run from root
+  dirs <- "tests/RUnit"  # Run from root
   test_suite <- defineTestSuite("rappor", dirs, testFileRegexp = "_test.R$",
                                 testFuncRegexp = "^Test")
   stopifnot(isValidTestSuite(test_suite))
diff --git a/analysis/R/unknowns_test.R b/tests/RUnit/unknowns_test.R
similarity index 95%
rename from analysis/R/unknowns_test.R
rename to tests/RUnit/unknowns_test.R
index 5efd7383..9478d775 100755
--- a/analysis/R/unknowns_test.R
+++ b/tests/RUnit/unknowns_test.R
@@ -27,12 +27,8 @@
 #     Both tests generate their own data.
 
 library(parallel)
-source("analysis/R/encode.R")
-source("analysis/R/decode.R")
-source("analysis/R/simulation.R")
-source("analysis/R/association.R")
-source("analysis/R/decode_ngrams.R")
-source("analysis/R/ngrams_simulation.R")
+library(rappor)
+
 alphabet <- letters
 options(warn = -1)
 
diff --git a/tests/analyze_assoc.R b/tests/analyze_assoc.R
index 5d78806f..5c50231e 100755
--- a/tests/analyze_assoc.R
+++ b/tests/analyze_assoc.R
@@ -49,11 +49,7 @@ if(!interactive()) {
   opts <- parse_args(OptionParser(option_list = option_list))
 }    
 
-source("../analysis/R/encode.R")
-source("../analysis/R/decode.R")
-source("../analysis/R/simulation.R")
-source("../analysis/R/read_input.R")
-source("../analysis/R/association.R")
+library(rappor)
 
 # This function processes the maps loaded using ReadMapFile
 # Association analysis requires a map object with a map
@@ -123,4 +119,4 @@ main <- function(opts) {
 
 if(!interactive()) {
   main(opts)
-}
\ No newline at end of file
+}
diff --git a/tests/assoc_sim.R b/tests/assoc_sim.R
index 3ff1e5df..c967e8bd 100755
--- a/tests/assoc_sim.R
+++ b/tests/assoc_sim.R
@@ -49,13 +49,9 @@ if(!interactive()) {
                 help = "Run simulation with uniform distribution")
   )
   opts <- parse_args(OptionParser(option_list = option_list))
-}    
+}
 
-source("../analysis/R/encode.R")
-source("../analysis/R/decode.R")
-source("../analysis/R/simulation.R")
-source("../analysis/R/read_input.R")
-source("../analysis/R/association.R")
+library(rappor)
 
 # Read unique values of reports from a csv file
 # Inputs: filename. The file is expected to contain two rows of strings
diff --git a/tests/compare_dist.R b/tests/compare_dist.R
index eb6521d3..4a99a328 100755
--- a/tests/compare_dist.R
+++ b/tests/compare_dist.R
@@ -36,6 +36,7 @@ if (is_main) {
                        positional_arguments = 3)  # input and output
 }
 
+library(rappor)
 library(ggplot2)
 
 # Use CairoPNG if available.  Useful for headless R.
@@ -47,12 +48,6 @@ if (library(Cairo, quietly = TRUE, logical.return = TRUE)) {
   cat('Using png\n')
 }
 
-source("analysis/R/read_input.R")
-source("analysis/R/decode.R")
-source("analysis/R/util.R")
-
-source("analysis/R/alternative.R")  # temporary
-
 LoadContext <- function(prefix_case) {
   # Creates the context, filling it with privacy parameters
   # Arg:
diff --git a/tests/gen_counts.R b/tests/gen_counts.R
index 769677c4..3602fd57 100755
--- a/tests/gen_counts.R
+++ b/tests/gen_counts.R
@@ -14,7 +14,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-source('analysis/R/read_input.R')
+library(rappor)
 
 RandomPartition <- function(total, weights) {
   # Outputs a random partition according to a specified distribution
diff --git a/tests/gen_true_values_test.R b/tests/gen_true_values_test.R
index e46d1e2b..894affed 100755
--- a/tests/gen_true_values_test.R
+++ b/tests/gen_true_values_test.R
@@ -2,7 +2,7 @@
 #
 # gen_reports_test.R
 
-source('analysis/R/util.R')  # Log()
+library(rappor)
 
 source('tests/gen_true_values.R')  # module under test