How determine numbers of trials to reach a good SEM

gvt · Post by **gvt** » Fri Jan 19, 2024 11:10 am

I need to determine how many trials or records are required to be sure that my SEM (Standar error of the mean) (or any realiability index) is real.

For example, I can do 3 records with the same conditions to get de SEM or CV, but its possible that I get low SEM only for chance. How be sure that is not for chance?

reason180 · Post by **reason180** » Sat Jan 20, 2024 4:20 am

FYI: Standard error of the mean is an estimate of the variability of a sample mean, if many samples were drawn from the same population. It's not a reliability index. (Perhaps you're thinking of 'standard error of measurement' rather than the standard error of the mean.)

It sounds like you want to assess whether the reliability is significantly different from zero. So you need to find either a significance test for SEm, or a tool to compute a 95% confidence interval for the SEm (then you would look to see if the interval excludes zero). Good luck in your search!

reason180 · Post by **reason180** » Sat Jan 20, 2024 3:27 pm

Also note that, like an effect size, the standard error of measurement (SEm) does not change systematically with sample size. So adding or subtracting trials will not tend to make the SEm more 'good' or less 'good.'

jamovi