Quote:
Originally Posted by Lune
[You must be logged in to view images. Log in or Register.]
This is correct, the distribution could be bimodal, steep, shallow, you don't know, and your rank/percentile depends on that. Think you need access to the data set, and with N=thousands, a computer/spreadsheet to calculate this, because it involves ranking the data.
Even if you had just the standard deviation you could assume a normal curve and estimate your percentile based on how many standard deviations you are from the mean, but you don't have that.
|
Agree.
Don't need the entire set, but you do need more than you have which is really just you. Look up Cochran's formula and you can calculate what sample size you'd need at any given confidence level. Then you might be able to look to a study for your data.
Ideally I think you'd want to use a study of Americans where weight was taken but not one related to weight loss/gain. Now is there a study like that where you can get the data set? No idea.