Bangladesh Sample Design

Sampling Plan and Timeline

Sample size for the Stigma in Global Context: Mental Health Study (SGC-MHS) is 1,500.  The total sample will be first divided into two parts—urban: 375 (25%) and rural: 1,125 (75%), according to urban/rural population distribution with slightly higher urban weightage given the relatively better awareness in the urban areas.

Both rural and urban sub-samples will then be distributed among the six administrative divisions, proportional to the populations of the divisions.

A multistage sampling procedure will be used for identifying the respondents.  First, two districts will be randomly selected from each division and, then, one thana will be randomly selected from each district, followed by random selection of two unions from each thana and then two villages from each union.  The divisional headquarters will be the urban centres where the surveys will be conducted.  The (rural) village selection is as follows, randomly at each stage:

The number of respondents from each division will be randomly picked from the villages identified in that division, allocating an equal number to each village.

While we have the listing of all administrative units at all levels including villages, household listing is not available.  The respondents will, therefore, be randomly selected from each village starting from a specific spot and moving to the right, picking every 6^th household.  This applies to both rural and urban areas.  However, in urban areas, a further two-way stratification will be made with reference to visible socio-economic conditions of the majority of the people in different parts of a particular urban area.

Replacement for refusal is automatic since any refusal(s) in any village or urban area will be replaced by continuing the process until the number of respondents allocated to a village or an urban area has been interviewed.

Only one respondent will be interviewed from each household. The aim is to interview men and women alternately. However, it is our experience that, in some households, women do not participate, even if the interviewers are women. Hence, although the male-female ratio should be about 50:50, in practice the proportion of females tends to be somewhat lower.  This is based on our past experience.

Assuming that there will be three Vignettes, one-third of the total sample, i.e. 500 respondents will be interviewed with reference to each Vignette, aiming to achieve 50:50 male-female ratios in each case, although in practice the proportion of women may be somewhat lower.

The interviewers will be appropriately and thoroughly trained.  Strict field supervision will be ensured and also a back-check will be conducted covering 10-20 per cent of the sample during and after the field work.

The questionnaire will be translated into Bengali and pre-tested to ensure clarity for both the interviewers and interviewees.

1022 E. Third Street, Bloomington, IN 47405 (812) 855-3841

Last updated: 6 September 2005
Comments: sgcmhs@indiana.edu
Copyright 2004, The Trustees of Indiana University
Copyright Complaints