# Ideas for sampling design

I have a data base of schools per district and the number of teachers per school. I want to select a sample where between 1 an 10 schools are selected in each district. Once a school is selected, the number of teachers selected should be proportional to the total number of teachers in the sample.

As we see, the districts are stratums, and the schools too.

EDIT: The maximum error of the proportion sholud be 0.035 and a 95% intervale confidence. I'm new to probabilistic sampling.

Is there any good tutorial or some advice to select the sample?

• Please tell us the purpose of this sampling. It might also be useful to explain why you are limiting the numbers of schools to this range from one through ten. – whuber Feb 9 '15 at 23:55
• It's a university project and the requirement of the teacher is that between 1 an 10 schools per district should be selected. The purpouse is to conduct a survey. – CreamStat Feb 10 '15 at 0:01
• Nobody conducts a survey just to conduct a survey: surveys are administered in order to learn things about the subjects surveyed and, through them, about the populations from which they were selected. What you want to learn determines, in part, what a good sample design will be. Other determinants include your expectations about the subjects' responses and the costs of administering the surveys. Since it is a school project, please consult our self-study wiki and apply its recommendations to how you frame this question. – whuber Feb 10 '15 at 0:06
• I would like to have more information too but I don't have information about the costs. The information I have is just what I wrote, and the maximum error of the proportion sholud be 0.035 and a 95% intervale confidence. I'm new to probabilistic sampling. – CreamStat Feb 10 '15 at 0:13
• Thank you for that additional information: you might want to include it in your question when you edit it. – whuber Feb 10 '15 at 0:45