Hi. I am undertaking to do a survey of Nigerian adolescent children aged from 10 - 17 years and to find out the way that they engage with and experience digital media technologies. I have decided to use cluster sampling or multistage sampling to arrive at my respondents: geographical zone - states - senatorial zones - local governments - towns - schools - classes. However my confusion and question is: how do I arrive at a sample size for the survey? Is the figure meant to be random? It should be justified, but how do I make this? What formula do I use and even so, how do I get total number of population?
Dear Uzuebgunam,
Infinite populations are very large and they are not feasible. Consequently, you should convert them into target populations by means of delimiting them based on population characteristics such as age, gender, level of education, socioeconomic status, etc. Subsequently, you can draw out samples of the right size from the population by choosing an appropriate sampling method.
Best regards,
R. Biria
Hi Reza
Thanks for your input. The problem is "what is the right size?" How do I justify any size that I decide to use?
Dear Emmanuel,
this link helps you:
http://www.surveysystem.com/sscalc.htm
first divide them into locations such as state and randomly select a few states. From states randomly selected, select the same number of respondents according to categories to ensure all category of the population are represented in the research
Firstly, start at the beginning. What sort of study is it?
qualitative or quantitative?
Quantitative observational descriptive, observational analytical, quasi-experimental or experimental?
If it is purely descriptive observational then the issue of sample size requirement devolves to "How accurate do I want to be in describing my variable of main interest?" and "are there any secondary variables that I am measuring/counting that will require larger numbers of participants to get the same degree of accuracy?" In the case of a proportional incidence are you satisfied with the 95% confidence limits about your likely observed proportion given by a 20, 50, 100, 1000.... subjects ?
As Tolu has said, the important issue in sampling for what you seem to want to do is representativity - ensuring that all sectors of the population are represented in the proportions in which they appear in the population. So you might want to stratify the individual participants according to age and gender (economic class etc. - whatever). In the case of which locations you sample (schools?) you might want to choose geographical zones and/or rural versus urban centres. I can't remember any rule for how many sites you should sample, but about 30% of each category would probably suffice. The issue of random sampling relates to how you wish to generalise your results. If you do not randomly sample your sites then your results cannot be taken to draw conclusions in any sites but those. If you don't randomly sample your participants, then you cannot use your results to draw conclusions about anybody but those study participants.
- Badges
- Science topic
- Similar topics
- Philosophy
- Ethics