The target study population for this cross-sectional study is people aged 18-35 years residing in a densely populated urban district. The sample size was calculated based on the number of people aged 18-35 years from the census data. As the study site is in an LMIC setting, there is no household mapping for the random sampling. Adding household mapping to the methodology would add both budget and time. It's a very small budget knowledge, attitude, and practice survey. What would be the best way to collect survey data in this scenario?
You said "The sample size was calculated based on the number of people aged 18-35 years from the census data." How did you do that? Did you consider a quantitative variable, continuous or at least a proportion (yes/no data) and a simple random or stratified random or cluster sample? If you are using simple random sampling, you should have considered how you were going to obtain the data before assuming you could use that technique. It sounds to me as if you need to do an area sample, using randomly selected plots (clusters) formed by geographically breaking down your urban area into blocks of some sort, which you would select at random. In single-stage cluster sampling you would census your 18 to 35 year-olds in each selected cluster/plot/block (whatever you want to call this level of unit). The subunits/elements of interest are your 18 to 35 year-olds.
I suggest you research cluster sampling. All data will then not be weighted equally, but you will only have to find the 18 to 35 year-olds in each selected cluster. (Your previous sample size calculation will not apply.)
PS - You could start your research with the following:
https://online.stat.psu.edu/stat506/lesson/7
At community level, official data are often non-existent, and those that exist are far less useful than information collected from the people themselves.
There are often local health workers who can advise you on which households have people in the target age group. Community nurses, for example, or traditional birth attendants. You can compile a sample frame from this information and then sample households. This often will yield more than one eligible participant per household, but you can use robust estimation of variance to account for clustering within household.
Another reason for using trusted community workers is that they can then introduce the survey (pay them for this!) and help to create a bridge of trust between participants and the research team.
Clearly, getting community participation should be the goal - I hardly have to remind you to involve them as much as possible in the design and conduct of your study. But I mention it because much LMIC research is carried out by white idiots jumping out of a 4x4 with clipboards.
Thank you for taking time to provide your insights. Engaging local community health workers is very important for research design and implementation at the community level. Local community youth clubs and youth advocates are improtant part of this research team.
- Badges
- Science topic
- Similar topics
- Psychology