Most researchers use 5% margin of error, and acceptable margin of error is said to be 4 to 8% but does it not depend on the proportion we take to use less than 4%?
Actually no. Your best choice is p=q=0.5 and z=1.96 or I.645 For details see the attached screenshot search book available in the z-library. Best wishes, David Booth
Besides the answer of David Eugene Booth , which i duly endorse there is also a role of the Power of the Test.
I would like to add that, from a statistical point of view, the role of calculating/determining the 'exact' sample size is often less important the quality of the sample. For instance, if the sample selection is biased and not representative, sample size is quite irrelevant. Hence, many journals will tend to focus on how the sample was collected, rather than on the formule that you used for calculating the sample size of a study.
Tewodros Getnet , my two cents about your original question (about the value of the margin of error): it can also depend on the actual money you have to collect the sample. I am working right now in a project where it is extremely expensive to collect the data (as it is in a dispersed area), and hence I am using a rather high margin of error (15%), since the "conventional" margin of error of 5% produces an extremely (costly) large sample for a proportion of p = 0.5 and a significance value of 1%, which seems also a little bit unreasonable given that the population size is N = 626, hence the use of finite population correction (cpf in the figures for each value of p and different significance values, some labels in Spanish, sorry, the project is for Latin America).
Tewodros Getnet , you may find this calculator useful, too:
Code Sample size calculators
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size