Document Sample
4 Powered By Docstoc
					Respondent-Driven Sampling: A New Form of
        Chain-Referral Sampling
 Center for Interventions, Treatment & Addictions Research
                     School of Medicine
         Wright State University School of Medicine
                3640 Colonel Glenn Highway
                    Dayton, Ohio 45435

• Survey Sampling.
• It has been a challenge for social researchers
to sample “hard to reach” or “hidden”
• Non-random sampling methods (e.g., chain-
referral and snowball sampling, and targeted
• Respondent-driven sampling (RDS).

                  Overview of RDS
1. Define target population
2. Conduct ethnographic research:
   - Better understanding of the target population
   - Identify seeds
3. Recruitment
   - Recruit seeds
   - Peer recruitment
   - Eligibility screening
4. Dual incentives in RDS recruitment process:
   1. The primary incentive
   2. The secondary incentive
5. Referral coupons
   - Impose rationing on recruitment
   - Collect recruitment process information
Example of a referral coupon:

6. Other necessary information to collect during
  - Network size (degree).
  - Network compositions.
  - Information on the social relationships between
    recruiters and recruitees.
7. Sample size

                  Advantages of RDS
1. Peer recruitment in RDS would minimize the issues
   about violation of subject confidentiality during
2. RDS helps reduce the effects of “volunteerism.”
3. The sample compositions will converge to reach
   equilibrium within a limited number of recruitment
   waves, independent of the characteristics of the initial
   sample (seeds).
4. RDS allows to estimate asymptotically unbiased
   population compositions.

5. RDS sample analysis enables to estimate unbiased
  sample statistics.
6. RDS sample analysis provides information for
  analysis of social structures.
7. RDS is cheaper, quicker, and easier to implement,
  compared with convenience sampling methods such
  as targeted sampling.

                Limitations of RDS
1. The target population size needs to be large enough,
   theoretically, infinitely large, to apply RDS.
2. There must exists a contact pattern among members
   of the target population.
3. Assume that participants recruit randomly from
   their personal networks.
4. Measurement errors are not taken into account.
5. Low ratios of referrals.
6. A successful RDS sample may not necessarily be
   representative of the target population.

                   Sample Analysis
1. Assess if the sample compositions have converged to
   reach equilibrium.
2. Assess the effect of volunteerism and masking.
3. Analyze the social structures of the target population.
4. Estimate asymptotically unbiased population
5. Estimate sample statistics from the weighted sample.
6. Computer program (RDSAT) for RDS sample analysis
   is free to download from website:
   Application of RDS among Stimulant
           Users in Rural Ohio
A sample of 249 participants was recruited between
June 2002 and February 2004 in three rural counties,
Ohio, U.S.
A total of 19 seeds were identifies: 8 of them did not
have any referral. A total of 230 respondents were
recruited for the study; 88.7% of the respondents
were linked to 5 seeds (see Table 1).

    Table 1. Seed’s Recruits, Seed-linked Recruitment Waves
                          and Recruits
                                                        Total No.
                 No. of         No. of Waves
 Seed ID                                               of Recruits
                Recruits          Linked
   001              2                  3                    5
   002              2                  4                    7

   003              4                  6                   19
   004              5                  6                   39
   006              1                  2                    2

   028              2                  4                    8
   037              5                  7                   60
   046              1                  1                    1

   152              3                 10                   64
   157              2                  6                   22
   166              2                  2                    3
Of the 19 seeds, 8 were “infertile” seeds (i.e., did not have any
referral ), thus not reported here. The total number of recruits
linked to the 11 seeds is 230.

Figure 1. Recruitment Tree Beginning with Seed 151

Heckathorn, D.D. 1997. Respondent-Driven Sampling: A New Approach to the
    Study of Hidden Populations," Social Problems, Vol. 44, No.2: 174-199.
Heckathorn, D. D., Broadhead, R. S., Anthony, D. L., and Weakliem,D. L.
    (1999).AIDSand social networks: Prevention through network mobilization.
    Sociological Focus, 32, 159–179.
Heckathorn, D.D. & Jeffri, J. 2001. Finding the beat: Using respondent-driven
    sampling to study jazz musicians. Poetics, Vol. 28: 307-329.
Heckathorn, D.D. 2002. Respondent-Driven Sampling II: Deriving Valid
    Population Estimates from Chain-Referral Samples of Hidden Populations.
    Social Problems, Vol. 49, No.1: ???-???.
Heckathorn , D.D., Semaan, S., Broadhead, R.S., & Hughes, J.J. 2002.
    Extensions of Respondent-Driven Sampling: A New Approach to the Study
    of Injection Drug Users Aged 18–25. AIDS and Behavior, Vol. 6, No. 1:
Salganik, M.J. & Heckathorn, D.D. 2004. Sampling and estimation in hidden
    population using Respondent-Driven Sampling. Sociology Methodology
    (in press).
Jichuan Wang, Robert G. Carlson, Russel S. Falck, Harvey A Siegal,
Ahmmed Rahman, Linna Li. Respondent-driven sampling to recruit MDMA
    users: A methodological assessment. Drug and Alcohol Dependence (in