Depends on what you want to do. How do you define accurate? Are you wanting to do some sort of statistical test or do you just want a fairly short confidence interval around your estimate? You need to be specific with what you are interested in.
Hi there!
I've a perhaps simple question: how can I estimate the number of measurements (cases) to obtain a accurate result?
Here an example:
I'ld like to know the average body length of a trout population in 100 ponds. The question is now, how many trouts of each pond do I need to measure. Indeed, measuring 100 per pond is better than 10. But is there somewhat like a statistical method to estimate the needed sample size per pond? I found some blogs dealing with reliability analysis, but they use it only for questionaries. Is there something similar for my case?
Thanks!
Depends on what you want to do. How do you define accurate? Are you wanting to do some sort of statistical test or do you just want a fairly short confidence interval around your estimate? You need to be specific with what you are interested in.
I don't have emotions and sometimes that makes me very sad.
I suggest you use a tool called G*Power (http://www.gpower.hhu.de/) that would allow you to determine the "required" sample size based on the desired effect size, alpha error probability, and power.
Hi and thanks!
Ok, I guess I formulated my question in a confusing context.. I try to estimate how many fish I need to measure to get an representative result.
In other words: what is my min. sample size to obtain a mean and standard deviation that is close (maybe 10% deviation) to the real values of the populations?
To actually be able to do this you need to be able to talk exact numbers. "Close" isn't something that you can plug into a formula. Do you want to be able to estimate the mean within 3 inches? What is your tangible goal. Those are the kinds of things you need to be able to answer.
I don't have emotions and sometimes that makes me very sad.
Hi Dason,
thanks again!
I guess I need to get my hands dirty with this topic before I come with further questions
Just get G*Power -- it does exactly what you ask for (taking in account Dason's comments, in a way).
Tweet |