Data Handling, Presentation, and Record Keeping

1
Data Handling, Presentation, and Record Keeping

• Jon Lundberg
• Nicole Aly
• With Special Guest Lecturer Liz Davis

2
Data Handling, Presentation, and Record Keeping

• Finding Data
• cleaned data
• Record Data Sources
• Databases
• Merging data bases
• Programs/Methodology
• Code-book
• Adjustments (e.g. variable/data deletion,)
• Equations
• Save
• Multiple locations
• Email, Flashdrive, RW-CD, etc.
• Silvia ? more space on network drive
• Set Deadlines
• Keep to them
• Takes Longer than you think

3
Data Reporting Example

• New York Times Article
• Calcium Study
• NO BENEFIT???

4
Econometrician Criticisms

• If you torture the data long enough, Nature will
  confess.
• Econometricians, like artists, tend to fall in
  love with their models.
• There are two things you are better off not
  watching in the making sausages and econometric
  estimates.

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
5
What to do?

• Discard the goal of objective inference
• Logical conclusion based on a set of facts
• Is it fact or opinion? Counterproductive
• Deciding if error terms are correlated. Could it
  depend on what I had for breakfast?

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
6
What to Report?

• Sensitivity Analysis
• Record inferences implied by alternative sets of
  opinions
• Show how an inference changes as variables are
  added or deleted from equation
• Do assumptions within the set lead to different
  inferences
• Work harder to narrow the set of assumptions

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
7
Fragility

• An inference is unbelievable if fragile
• Can it be reversed by minor changes in
  assumptions?
• Can it stand up to researchers opposite opinions?
• Do your own detailed sensitivity analysis
• Report mapping from assumptions to inferences
• Anticipate opinions of consuming public

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
8
Mapping Problems

• Space of assumptions is infinite
• Example Measuring effects of fertilizer on crops
• If you include light level, what about rainfall
• If you include rainfall, what about temperature
• If you include temp., what about soil depth
• If you include depth, what about soil grade
• And so on, and so on, and so on, and so on
• YOU GET THE PICTURE!?

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
9
Important Question

• Coefficient is negative and you think it should
  be positive. You find another variable to add to
  make it positive. Have you found evidence that
  the coefficient is truly positive?
• Think about what you would do before examining
  the data.

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
10
Murder Rate Example

• Simple Regression Results Each additional
  execution deters 13 murders, S.E. 7
• Is this conclusion fragile?
• Different viewpoints different subsets

Source Leamer, Edward E., Lets Take the Con
Out of Econometrics
11
Data Confidentiality

• What types of data are private or confidential?
• What laws and regulations protect the privacy of
  data used by researchers?
• What responsibilities do researchers have to
  protect the data?
• What do we mean by ethical and responsible use of
  data?

12
Which data might be sensitive?

• Social Security Numbers
• Often SSNs are one of the few identifiers
  available to match individuals across data sets.
• Names and addresses
• Individuals, companies, public officials,
  students
• Geographic identifiers
• County, city, census tract, geocode
• Employment and earnings
• Both employers and employees concerned
• Health data, government benefits, tax data

13
Laws and regulations

• Census Bureau confidentiality rules
• Minnesota Government Data Practices Act (MGDPA)
• HIPAA Privacy Rule
• Confidentiality agreements with government
  agencies or private companies or with survey
  respondents

14
Protected information, for purposes of this
agreement, includes any or all of the following

• Private data (as defined in Minnesota Statutes
  13.02, subd. 12), confidential data (as defined
  in Minn. Stat. 13.02, subd. 3), welfare data (as
  governed by Minn. Stat. 13.46), medical data (as
  governed by Minn. Stat. 13.384), and other
  non-public data governed elsewhere in Minnesota
  Government Data Practices Act (MGDPA), Minn.
  Stats. Chapter 13
• Medical records (as governed by the Minnesota
  Medical Records Act Minn. Stat. 144.335)
• Chemical health records (as governed by 42 U.S.C.
  290dd-2 and 42 CFR 2.1 to 2.67)
• Protected health information (PHI) (as defined
  in and governed by the Health Insurance
  Portability Accountability Act HIPAA, 45 CFR
  164.501) and
• Other data subject to applicable state and
  federal statutes, rules, and regulations
  affecting the collection, storage, use, or
  dissemination of private or confidential
  information.

15
Responsibilities of Data Users

• 1) Know the rules Ignorance is not an acceptable
  defense.
• 2) Understand the purpose and objectives of data
  privacy.
• 3) Protect data from unauthorized access or
  disclosure.
• 4) Do not report data that can identify
  individuals or specific companies unless it is
  public information.
• 5) Destroy confidential data when no longer
  needed.

16
Ethical Issues in Reporting and Presenting
Results

• Potential disclosure of individual or sensitive
  data.
• Cleaning data and eliminating outliers
• Reporting findings that support ones hypotheses
  and suppressing those that dont.
• Statistically insignificant results.
• Subgroup analyses and multiple hypothesis tests.
• Data mining.

17
Questions?
18
(No Transcript)