PAD 637: Lecture

1
PAD 637 Lecture 4 Data Collection and
Social Cognition 09/29/04 Rinata Shamshetdinova
2
Techniques for social data collection/1

• questionnaires
• roster vs. free call
• free vs. fixed choice
• ratings vs. complete ranking
• interviews
• face-to-face
• telephone
• observations
• Planting an observer in a room
• Random appearance, each time over a limited period

1
3
Techniques for social data collection/2

• archival records
• structured data sets
• unstructured data sets (coding required)
• experiments
• planting rumors
• small world
• others
• social cognitive structures
• ego-centered
• small world
• diary

2
4
Boundary Specifications Sampling

• 2 approaches to boundary definition
• Nominalist approach analyst self-consciously
  imposes a conceptual framework constructed to
  serve his or her own analytic purposes.
• Realist approach the boundaries of the social
  entities are defined by actors themselves. Social
  entity exists as a collectively shared subjective
  awareness of all, or at least most, of the
  actors.
• Sampling
• Relevant sampling units?
• Estimation of network properties based on
  sampling units
• Snowball network sample

3
5
Information accuracy

• technical aspect - the accuracy of data
  collection
• What data do we need?
• cognitive aspect the accuracy of informants
• Krackhardt behavioral measures of interaction
  are not very closely related to participants
  self reports of the same interactions often
  recall is being used as a surrogate of behavior.
• Bernard, Killworth and Sailer about 52 of what
  informants said was wrong.
• Freeman, Freeman and Romney people's recall of
  their interactions with others is systematically
  biased toward what is normal and/or logical
  recalling important people more then others.

4
6
Cognitive Social Structures

• CSS
• sender
• receiver
• perceiver
• Rijk person k thinks that person i approaches
  person j for help and advise
• Characteristics
• Amount of information in CSS gt that in
  traditional SS ? problems in data collection
• No assumption is made about the perceptions
  being correlated with each other.

5
7
Useful bits/1

• Informants who usually kept records of their
  behavior were not more accurate than those who
  did not. (BKS)
• Slightly obtrusive observation has no noticeable
  effect on informant accuracy. (BKS)
• Telling people in a group that they are expected
  to get more accurate in repeated experiments over
  time produces no significant improvement in
  accuracy of reporting communications. (BKS)
• Asking people who do you like? produces about
  the same answers as asking them who do you talk
  to? (BKS)
• Asking people about the significance or
  importance of their interactions with others is
  of little use, since it produces no better
  results than simply asking them who they talked
  to. (BKS)

6
8
Useful bits/2

• In general, the more recent the time of the
  window over which informants report their
  communication, the more accurate they tend to be.
  However, only 6 of informant accuracy can be
  accounted for by this factor. (BKS)
• Over all data sets, people can recall or predict
  less than half of their communications, measured
  either on amount or on frequency. (BKS)
• For publicly available information there was
  both agreement and accuracy. For questions where
  the answers were not public, there was very
  little agreement, and hence high inaccuracy.
  (Young Young)
• People represent enduring patterns of
  interaction more accurately then individual
  instances of behavior (FR)

7
9
Question 1

• Does agreement presupposes accuracy, especially
  when we deal with a closely-knit network? Under
  what conditions would it be true?

8
10
Question 2

• Accuracy vs. problems in data collection and
  treatment should we, at this stage use Cognitive
  Social Structures in our Empirical Exercises?

9
11
Question 3

• Under what circumstances can random sampling be
  used???

10
12
(No Transcript)