Talk:Statistics/Testing Data/z-tests

This article is highly unsatisfactory to explain its topic!130.89.219.179 23:21, 2 October 2005 (UTC)

The example "Is The Machine Working Correctly?" is seriously flawed. A quick inspection of the data shows that the standard deviation is nowhere near the 0.05 value quoted. The z test does not take into account the sample standard deviation since it assumes that the standard deviation of the population is known. Which test can be used to determine the likelihood of getting this sample standard deviation?

sample size for non-normal populations
The n=30 rule is over-simplistic. The answer really depends on whether there are outliers and/or skewness in the population distribution. With a symmetric, but non-normal distribution, n values smaller than 30 will suffice. With even moderate skewness, n>30 will be necessary. With outliers, n may have to be hundreds or even thousands before the CLT has converged sufficiently.