Project D: Analysis and Design of Alumni Surveys
Investigator: Judy Cole, Associate 
   Vice President for Alumni Relations, Carnegie Mellon
Team: Yi Jiang, Zhanwu Liu

=====================================================================

Notes from first presentation (Zhanwu only)

Alumni & advancement

  "Promotes the university brand name reputation"

  "Responsible for external relations and development. Offers
   profiles, gift giving and speakers." 

  "All of the units of University Advancement work together to bring
   recognition and resources to the university."

  "Overview, introduction, and duties profile for the staff overseeing
   Alumni Relations, Development Services, Public Relations and
   Creative Services, ..."

  "The mission of the Division of University Advancement is to develop
   greater understanding and support for University goals, objectives,
   programs, and activities among many publics."

Clients
  Judy Cole
  Christian ..

Ultimate goal: panel survey design

Data

 - Alumni database 77K records

 - previous surveys
     2002    email, mail, web
     2004    email, some mail
     2006    email only

 - can get
     - database
     - lists of people in the 04 and 06 surveys

 - need conf agreement at some point

Questions

 - bias in email survey

 - Design panel survey / stratification
   (try sim sampling in existing database to compare to 'pop'?)

 - ID "active" alumni, estimate "how many alumni"...
     - what are the characteristics of "responsive" alumni
         - survey respondents
         - gift givers
         - participators or committee members, etc.

Difficulties

 - Messy data

 - what is the response variable

 - optimal sampling method?

Technical

 - SAS?

 - cap/recap (varying "attractivity" across individuals might suggest
   Rasch e.g.)

Their proposed analyses/report:

1. Analyze the bias of email-based survey
   - How people with email different from people has no email address
     in file. 

2. Analyze the involvement in CMU advancement 
   - What is the difference between alumni who responded twice, once
     and none to previous surveys. 

3. Simulate sampling using different strata
   - Which stratifying scheme represents the population?

4. Capture/recapture analysis to analyze the population size?

5. Design the panel based on all above.

===============================================================