Resolved: Confidence interval for non normal distribution
In the lesson we calculate the confidence level for a sample taken from a population which is normally distributed (salaries).
The lector also says - "if you wart with a sample, which is large enough you can assume normality of sample means" (CLT)
My question is, how should I calculate confidence interval for a non-normally distributed data.
Option 1.
Take ONE sample, calculate the confidence interval using the forumula mean_of_this_ONE_sample +/- margin error
Option 2.
Take MANY samples. calculate the mean of each sample and build a new normally distributed set of data (sample means like in the CLT). Next, calculate mean on the new data set and finally use it in the equation for the confidence interval using the formula mean_of_means +/- margin error
Which option should I use?
Hi R K,
In most cases, you will be able to have only 1 sample, so that's the typical way to approach this task. You can use the T-table, which incorporates a margin of uncertainty because you work with sample data.
Best,
Ned
