Github Account: https://github.com/mingzehuang

acpc: Area Classification on Dimension Reduced Micro-level Census Data

Acknowledgment: This package is inspired by several similar assignments by Irina Gaynanova when I’m taking Statistical Computation at Texas A&M University as a graduate student.

The R package acpc implements robust PCA, then sparse PCA, then K-means classification for Micro-level census data. The original idea was inspired by

Vickers D. and Phil R. (2007). “Creating the UK National Statistics 2001 output area classification”. J. R. Statist. Soc. A.

The input data matrix is expected to be n rows, p + 1 columns data frame, which is commonly used in census data. The first column contains name IDs for all n observations, the rest p columns are features. The column names for features are feature IDs. All features are supposed to be numeric so that robust sparse PCA and K-means can be processed meaningful.

The package will omit those observations with missing values.

Output list includes the cluster labels for all (non-missing) observations corresponding to observation names; the scores for clustering centers; the scores for observations; the loading coeffcients for corresponding principle components and features.

Installation

devtools::install_github("mingzehuang/acpc")

Example for function acpc()

library(acpc)
# Demographic data for each county in US
demo_data = readRDS(url(
"https://shiny.rstudio.com/tutorial/written-tutorial/lesson5/census-app/data/counties.rds",
  "rb"))

# Set number of clusters
K = 3

# Classification outputs
acpcresult <- acpc(demo_data, K)
acpcresult$Y # Cluster label for each observation.
acpcresult$center # Scores for cluster centers.
acpcresult$U # Scores for observations.
acpcresult$V # Loadings for corresponding principle components and original features.

Example by default data for function visualization()

Don’t run this chunk in Rmd/md file if you don’t like warning. Shinny App requires continuous serving, you will see warning if you stop serving.

Please make sure you connect to internet since the default data is online.

You should run it in your console.

# Shinny App visualization (only works for two principle components case!)
if(exists("data_for_acpc")) {
rm(data_for_acpc)
} # Clean your global environment to avoid name conflict!

# Generate App for default dataset

visualization()

Use your own data for function visualization()

Don’t run this chunk it in Rmd/md file if you don’t like warning. Shinny App requires continuous serving, you will see warning if you stop serving.

You should run it in your console.

# Generate App for your own dataset

your_own_dataset = readRDS(url(
"https://shiny.rstudio.com/tutorial/written-tutorial/lesson5/census-app/data/counties.rds",
"rb")) # Load your own dataset

data_for_acpc = your_own_dataset # Rename your own dataset as "data_for_acpc"!
visualization()