Determinants Of Income Variability Across Population Clusters - PowerPoint PPT Presentation

1 / 6
About This Presentation
Title:

Determinants Of Income Variability Across Population Clusters

Description:

Are the determinants of income same across population clusters? ... Neural Networks can model non linearity, and interaction terms ... – PowerPoint PPT presentation

Number of Views:20
Avg rating:3.0/5.0
Slides: 7
Provided by: meetao
Learn more at: http://www.mscs.mu.edu
Category:

less

Transcript and Presenter's Notes

Title: Determinants Of Income Variability Across Population Clusters


1
Determinants Of Income Variability Across
Population Clusters
  • Chonkar, Hemendra
  • Gautham, Arthi
  • Ghosh, Anirban
  • Oberoi, Meeta

2
INTRODUCTION
  • Concept
  • Model the determinates of median household income
  • Process
  • Identify homogeneous population clusters
  • Model the determinates of income within these
    clusters
  • Critical Questions
  • Are the determinants of income same across
    population clusters?
  • Policy Implication If not same, then targeted
    policy initiatives appropriate
  • How to measure magnitude of determining factors
  • Neural Networks can model non linearity, and
    interaction terms
  • Linear Regression Models standard econometric
    technique
  • Analysis Implication If the two analysis forms
    produce different results, potentially recommend
    using neural networks more in econometric
    analysis

3
Process
Interpretation and Write up
Regression Analysis
Neural Networks Analysis
Cluster Interpretation (11.18.2004)
We are here
Clustering (11.15.2004)
Data download and Preprocessing (11.3.2004)
Proposal Approval
Done
Data Source Identification (10.25.2004)
Concept Creation (10. 23.2004)
4
A Preliminary Cluster Solution
  • Preliminary Clustering Solution on standardized
    data
  • Hierarchical clustering used (Agglomerative)
  • Clustering Tool Matlab
  • Next Step (The ART of cluster analysis)
  • Interpret the cluster means to see if reasonable
    clusters are formed
  • Decide on appropriate number of clusters
  • Compare results with other clustering techniques

5
Challenges
  • Data Download
  • Over 30,000 observations and 30 attributes
  • Can download only 7000 observations and 1
    attribute per run
  • Preprocessing, data transformation and data
    warehousing
  • Attribute Selection
  • Reference Paper and previous research experience
    used as a guide
  • Correlation Analysis
  • Tool Selection
  • Large Dataset not supported by Weka
  • Currently using
  • Access, Excel Data Warehousing
  • Matlab and SPSS Data Analysis

6
Project Details
  • Data Source U.S Census Data from Summary File
    3, that contains data related to household
    income.
  • http//factfinder.census.gov/
  • Data Collected at Zip Code level (30,000 data
    points)
  • Reference Paper Regional Variables in Median
    Family Income by J.H. Chesnut
Write a Comment
User Comments (0)
About PowerShow.com