Question on Standard Error

#1
If I need to compute the standard error for a bunch of students' test averages, and there are 20 students, each of whose score is itself an average of 8 tests, will my N for SEM=sdev/SqRt(N) be 20? Or will it be 160? In other words, if my bar graph plots the mean of 20 students' averages, but each of those averages is itself an average of a bunch of tests, what is my N?
 
#2
Hi, I have the exact same question and would really appreciate an answer.

I am quite sure that it is 20 in your example. But the point of confusion for me is with the word sample - I understand that we refer to those 20 students as a sample to represent the population from which they were taken. But actually we are also gathering a sample of raw test scores from each participant, thus we end up with 20 samples for an N of 160.
 

rogojel

TS Contributor
#3
If I need to compute the standard error for a bunch of students' test averages, and there are 20 students, each of whose score is itself an average of 8 tests, will my N for SEM=sdev/SqRt(N) be 20? Or will it be 160?
hi,
I think it will depend on what you use as a value for the stdev. If it is the standard deviation of the averages then your N is 20. If it is the standard deviation of the individual 160 values then it is 160.

Actually the two values should br pretty close to each other - "equal" for mathematicians.

regards