This is where multi-stage sampling can be very useful. We have used it in situations where we want to recruit children under the age of 5 living in remote areas in Africa.
We would begin by selecting health clinics at random in the districts where we were surveying. We would visit each clinic and get a list of the villages in their catchment areas (this information was more likely to be accurate if you asked the clinic). We would then visit a random sample of villages with a health worker and get them to show us all the houses with young children. Finally, we would randomly select houses.
In this way, you build you sample frame as you go.
Remember, however, that such samples cause clustering in the data, so you have to use robust variance estimation in your statistical methods. In Stata, these methods are built in, either using the -robust- option or by setting the data up as survey data using the -svyset- command.
This is where multi-stage sampling can be very useful. We have used it in situations where we want to recruit children under the age of 5 living in remote areas in Africa.
We would begin by selecting health clinics at random in the districts where we were surveying. We would visit each clinic and get a list of the villages in their catchment areas (this information was more likely to be accurate if you asked the clinic). We would then visit a random sample of villages with a health worker and get them to show us all the houses with young children. Finally, we would randomly select houses.
In this way, you build you sample frame as you go.
Remember, however, that such samples cause clustering in the data, so you have to use robust variance estimation in your statistical methods. In Stata, these methods are built in, either using the -robust- option or by setting the data up as survey data using the -svyset- command.
Ramadan. I link "sampling frame" with the research context that produces your dataset. Do you give the same sense to it? And second point, your tarjet is "the best method to obtain a representative sample". I think we need first to agree about what is a "representative sample, a question I can only answer if variable distribution is well known apriori -assuming a univariate case-.
Thank you very much for your contributions. The issue in case of online survey studies.. where the sampling frame is undefined e.g. Coca Cola drinkers in Middle east countries. in this case the population is existed but the frame is undefined !!
i have same problem as you. i would appreciate if you share your finding to me also. i dont have sample frame . i have only the size of population and also size of my sample but no contact info and i want to do online survey as well.
Elham, do you mean that you do not have any variable to be measured, neither some unit to describe the variable size? I understand that you only have a population of your interest, but no problem nor question of research that interests you. Am I wrong? Emilio