SAS/INSIGHT Demonstration - Exploratory Data Analysis (EDA) techniques

1
SAS/INSIGHT Demonstration - Exploratory Data
Analysis (EDA) techniques

• David Lassman
• Office of the Actuary
• X7933
• David.Lassman_at_cms.hhs.gov

2
Introduction

• Apply exploratory data analysis techniques (using
  SAS/INSIGHT) to improve outlier detection
• Demonstration
• Datasets
• Census Bureau
• Service Annual Survey
• Industry estimates
• American Hospital Association
• Annual hospital data
• Baseball data 1986 statistics
• Hitters

3
Exploratory Data Analysis(EDA)

• A set of tools for finding what we might have
  otherwise missed in a set of data (Tukey, 1977)
• Determine which cases are unusual with respect to
  the bulk of the cases and to follow-up those
  cases.
• Limited time frame

4
Data Review using Graphical Data Analysis-Benefits

• Allows you to see the big picture
• Research/explain data trends
• Find patterns in your data

5
Techniques

• Box plot
• Location, spread, and shape of a distribution
• Scatter plot
• Fitting
• Ordinary least squares regression
• Transformations
• Log
• Square root
• Other

6
Demonstration datasets - Service Annual Survey

• Collect financial data
• Revenue
• Expenses
• Service Industries include
• Personal Services
• Business Services
• Health Care and Social Assistance
• Trucking and Warehousing

7
Health Care and Social Assistance (NAICS 62)

• Physicians - NAICS 6211
• Dentists - NAICS 6212
• Other health care practitioners - NAICS 6213
• Chiropractors NAICS 62131
• Optometrists NAICS 62132
• Outpatient Care Centers NAICS 6214
• Home Health Care NAICS 6216
• Other Ambulatory Services NAICS 6219
  examples blood/organ banks, health screening
  services, hearing services
• Hospitals NAICS 622
• Nursing and Residential Care Facilities NAICS
  623

8
Health Care and Social Assistance (NAICS 62) cont

• Data items
• Total Revenue
• Source of payment
• Medicare
• Medicaid
• Other Government
• Patient (out-of-pocket)
• Patient from family NAICS 623 only
• Patient from social security benefits NAICS 623
  only
• Private insurance
• All other patient care
• All other services
• Total Expenses

9
-----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes-----
Variable Type Len Pos Format Informat Label
17 Description Char 47 135      
2 ITEM Char 24 110 24. 24. ITEM
1 NAICS Char 6 104 6. 6. NAICS
3 Tax Char 1 134 1. 1. Tax
15 ch_00_99 Num 8 88 PERCENT6.2   ch_00_99
14 ch_01_00 Num 8 80 PERCENT6.2   ch_01_00
13 ch_02_01 Num 8 72 PERCENT6.2   ch_02_01
12 ch_03_02 Num 8 64 PERCENT6.2   ch_03_02
11 ch_04_03 Num 8 56 PERCENT6.2   ch_04_03
16 ch_99_98 Num 8 96 PERCENT6.2   ch_99_98
10 r1998 Num 8 48 DOLLAR12.   r1998
9 r1999 Num 8 40 DOLLAR12.   r1999
8 r2000 Num 8 32 DOLLAR12.   r2000
7 r2001 Num 8 24 DOLLAR12.   r2001
6 r2002 Num 8 16 DOLLAR12.   r2002
5 r2003 Num 8 8 DOLLAR12.   r2003
4 r2004 Num 8 0 DOLLAR12.   r2004
Service Annual Survey -
10
AHA hospital dataset
-----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes-----
Variable Type Len Pos Format Informat Label
6 ADMTOT Num 8 32     ADMTOT
5 BDTOT Num 8 24     BDTOT
7 BIRTHS Num 8 40     BIRTHS
2 EXPTOT Num 8 0 DOLLAR27.2 DOLLAR27.2 EXPTOT
1 ID Char 10 48 10. 10. ID
8 MNAME Char 30 58 30. 30. MNAME
4 NPAYBEN Num 8 16 DOLLAR27.2 DOLLAR27.2 NPAYBEN
3 PAYTOT Num 8 8 DOLLAR27.2 DOLLAR27.2 PAYTOT
9 bdgrp Char 18 88      
11
-----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes----- -----Alphabetic List of Variables and Attributes-----
Variable Type Len Pos Label
19 assists Num 8 112 Assists
5 atbat Num 8 0 Times at Bat
12 atbatc Num 8 56 Career Times at Bat
22 batavg Num 8 136 Batting Average
23 batavgc Num 8 144 Career Batting Average
20 errors Num 8 120 Errors
6 hits Num 8 8 Hits
13 hitsc Num 8 64 Career Hits
7 homer Num 8 16 Home Runs
14 homerc Num 8 72 Career Home Runs
2 league Char 1 166 League
1 name Char 14 152 Hitter's name
4 position Char 2 170 Position(s)
18 putouts Num 8 104 Put Outs
9 rbi Num 8 32 Runs Batted In
16 rbic Num 8 88 Career Runs Batted In
8 runs Num 8 24 Runs
15 runsc Num 8 80 Career Runs Scored
21 salary Num 8 128 Salary (in 1000)
3 team Char 3 167 Team
10 walks Num 8 40 Walks
17 walksc Num 8 96 Career Walks
11 years Num 8 48 Years in the Major Leagues
Baseball -
12
SAS/INSIGHT

• Interactive tool for exploring and analyzing
  data.
• Graphs include box plots, histograms, and scatter
  plots
• Dynamic - Click on points to identify
• All graphs and analyses are linked.
• Brush observations in one window and they are
  highlighted in all windows.
• Color a point in one graph and it receives the
  same color in all displays.
• Exclude an observation from calculations, and all
  analyses recalculate automatically.

13
Demonstration

• Invoke insight
• Type insight in the command box
• P-menu solutions-analysis interactive data
  analysis
• PROC insight