## Question 339:

Asked on August 12, 2008

Tags:
Sample Size ,
Margin of Error

1

## Answer:

No answer provided yet.

- It looks like you are sampling from many subgroups of populations. If you want to make inferences about facilities, then you'd need to sample from each of them proportionally. If you want to then make inferences about the divisions within a facility, you'd have more subgroups and would need to sample proportionally from each.
- My advice is to work backwards. What is the highest number of people you could likely sample given time constrains, money and getting access to these people. Usually this is a number that's pretty fixed. Then try and spread that across a reasonable number of subgroups.
- Before you consider sample size, you should consider what it is you're hoping to measure. You stated people will respond to agree/disagree questions. I'd recommend coding the agree to disagree into numbers from 1 to 5 where higher numbers represent higher disagreement, if you sampled 20 people from one facility and had a mean response of 2.2 and standard deviation 1.4, you'd have a margin of error around your mean or +/- 30%. What level of precision are you looking for? +/-5%? That requires a larger sample. As your need for precision increases, your sample size will increase. But what will you do with the mean of 1.4, does that have meaning?
- As long as you have one response and it is the same type of response per person, all methods (written, interviews) of conducting the sample have the same sample size calculations.