2006 China Sexuality Survey Sampling

Suiming Pan, William Parish, Yingying Huang

     Sample. The 2006 Sexuality Survey of China used a stratified, multi-stage cluster sample [1]. In each of the urban and rural strata, official administrative units were selected with probability proportional to size (PPS) – larger administrative units were more likely to be picked, based on a random number generator. Sampling proceeded through the following stages: (1) In the urban stratum 75 primary sampling units (PSUs) were selected from a list that included all urban districts (in larger cities) as well as whole towns and small cities. From all rural counties, 47 counties were picked. (2) A single administrative neighborhood (jiedao) was selected from each urban PSU and a single administrative subunit – be it township (xiang) or town (zhen) – from each rural PSU. (3) Two resident’s committees (juwei hui) were selected from each urban neighborhood and one village (cun) level unit from each township/town. (4) At the lowest level, individuals 18-60 years old from the local household register (for both permanent and temporary residents) were listed. In urban neighborhoods lacking a complete register of migrants, the residents list of residents was supplemented by local registration officials providing both an estimate of the number of unregistered migrants and sites (peddlers, small stores, and construction sites) where migrants worked. These sites were enumerated by the survey team and then added to list of residents. This complete list was sampled by using a random number generator to pick a starting point and then picking every nth person after that point (with the nth gap chosen to produce the requisite number of eligible persons for that locale).

      Participants were interviewed not at home but in a nearby meeting place – hotel, school, neighborhood office – with family members absent. The questionnaire was in a laptop computer, with a same-sex interviewer controlling the computer initially followed by the participant controlling the computer during the last, more sensitive part of the interview. In urban areas, the proportions of men needing help in using the computer during the sensitive portion of the questionnaire was “never” (54%), “sometimes” (42%), and “constantly” (4%). In rural areas, these percentages were 40, 50, and 11%. The study procedures were approved by a Renmin University human subjects committee, and prior to beginning the interview participants gave informed consent.

      In the initial sample of 10,203 individuals, 26% could not be contacted at the address where they were registered – with many of them having migrated elsewhere [].  This left 7,553 individuals who had the potential to be interviewed. Excluding refusals, computer glitches, and faked interviews, 5,404interviews were completed and available for analysis. Or, compared to the 7,553 individuals who could potentially be interviewed, the response rate was 71.5%.

      Some questions were not answered – because of a few participants breaking off the interview before getting to questions near the end of the interview, because of computer glitches, and because of errors in data cleaning. The implications for analysis samples were as follows: First, the cross-national analysis sample for all men 18-49 years of age was reduced from an initial 2,231 men to 2,222 men. Second, through deletion of observations lacking information on either outcomes or risk factors, the sample of urban men 18-60 years of age was reduced from an initial 2,079 to 2,066 in the analysis of risk factors for FSW visits and to 2,068 for the analysis of risk factors for sexually transmitted infections (STI).

Endnotes:

1.       Levy PS, Lemeshow S. Sampling of populations: Methods and applications. New York: Wiley; 1999.

2.      Landry P, Shen M. (2005). Reaching migrants in survey research: The use of the global positioning system to reduce coverage bias in China. Polit Anal 2005; 13:1-22.