Hi all,
I had a tough time understanding these 3 statistics and let me share what I understood with you guys.
- The first thing we should know is that we will always be dealing with samples from a large population. For example, say we want to know the average height of a Japanese woman. We will measure the height of n=100 Japanese women and we will draw some conclusions about the entire Japanese women population.
- Now, standard deviation (SD) of the 100 measurements will give us how our measurements vary around mean of the measurements.
- the standard error (SE) (of the mean) will tell us the variability of our computed mean. I repeat; SD gave the variability of data around the mean and SE gives the variability of the computed mean, not the data around it.
- finally, the confidence interval (CI) give us where the real mean of the large population might lie.
- Now, how do we compute these 3 values?
- SD is computed directly from our measurements.
This is how you get it.
is ns^2 (because for two it will be s1^2 + s2^2)
The variance of T/n (mean) must be 1/n^2 * (ns^2) = s^2 / n
Thus, the standard deviation of T/n = s/(sqrt(n))
- CIs are computed from the SE.
- Say, we need a 95% CI, then
1.96 comes from 0.975quantile of the normal distribution -- (100-95)/2%=0.975