Relational and Network Data

1
Relational and Network Data

• Mike Ryckman

2
Overview

• Relational Data Concepts
• Network Analysis Data
• Events Data

3
Relational vs. Flat Data

• Academics are used to flat data
• Relational data is more common elsewhere
• Flat data Excel/Stata
• Relational data Collection of related flat
  tables
• Relational data structures are trickier Well
  look at an example

4
Running Example Legislative Action

• Say we are interested in studying how legislators
  work together to pass a bill
• We interview the legislators asking about
  different pieces of legislation and ask who
  worked to pass the bills
• We also record some secondary data how did they
  help, did the bill pass, etc.
• This type of analysis could tell us a lot about
  the legislature who are the major players? how
  and when do they work together?

5
Data setup

• Say you want to record bill participation
• Excel
• What if someone participates on two bills?

6
Option 1

• Repeated data what if you mistype Erin?
• Wasted space imagine doing this 10,000 times

Option 2 Very Common

• Bad for searching how do you see who was on a
  specific bill?
• Bad for combining data how do you count bill
  participation?

7
Relational Data Structures

• Instead of one table what about using 3?
• Table 1 holds your people, table 2 holds your
  bills, table 3 is bill participation

8
Advantages

• By storing data relationally, we can do anything
  we want with it its flexible
• You minimize data storage and entry one table
  can hold all your information on your legislators
  so you only have to enter it once
• It is VERY efficient computers (and computer
  nerds) love it
• You can generate flat files from relational files
  in one step it is difficult to do the opposite

9
Notes

• You generally need flat data for statistics (this
  is why academics like it)
• Flat data is GENERALLY less efficient
• It is better to know what this stuff is then know
  how to use it
• Remember context choose wisely

10
Network Analysis Data

• Network data lends itself well to relational data
  structures
• Generally, we are interested in producing a
  matrix of values

11
Back to Our Example

• Think in terms of the goal
• There are a number of matrixes we could create
• N x N matrix of legislators
• diagonal the number of bills they worked on
• off-diagonal number of bills two legislators
  worked on together
• Legislators x Bills
• Singular indicators for whether the legislator
  did or did not participate on the bill

12
Two Matrixes
NxN Matrix
Using matrix algebra XX
13
Generating the Table

• Getting that table is not always easy
• Most of the time your data will not be structured
  this way
• Relational databases have a special type of query
  that will create this (a crosstab or transform
  query) but most have limited field counts (about
  250)
• Other options are available

14
Setting up your data

• Ultimately, it comes down to how you setup your
  data
• In a relational db, you can have 3 tables
  legislators, bills and bill participation
• You could also build the table manually, or use
  some other tool

15
More Notes

• There are many ways of structuring and using your
  data
• Some are easier than others depending on the
  situation and your knowledge base
• Dont get bogged down in the details there are
  always people around who can help you with the
  data

16
Events Data

• Events data sets (nowadays) are machine coded
• A computer program reads a news feed and
  determines who, what, when, where and how an
  event occurred
• The VRA Events data has about 10 million records
  from 1990 2004
• It includes both domestic and international events

17
Events Have

• A source and a target (for domestic events source
  target)
• Level information (government, businesses,
  organizations, groups, etc)
• A date specific to the day

18
Notes on VRA Events Data

• It is bias toward the US about 3 million
  records or so have the US in them
• It covers a HUGE range of activities wars,
  meetings, animal attacks, etc.
• The data are very flexible and we can create
  matrixes of all types out of it