I have read that it cant properly work in a non normal distribution and that it does. What is it
thank you!!
It depends on what you mean by “it doesn’t work”. For a non-normal distribution, you don’t have the same cutoff points for the confidence intervals that you do in a normal. But in many distributions, the SD is still useful - like a T, or binomial, or hypergeometric (just off the top of my head). In fact, for large enough sample sizes, you can use a normal approximation for the binomial.
The tables that you have for the normal distribution won’t work if you’re not dealing with a non normal distribution. Maybe that’s what you read?