Reliability of Differences Between Correlated Means (Nណ)
We use t test when we have small samples.
- Differences between means will be normally distributed about the actual mean difference of zero (0) if we draw samples randomly from the same population.
- As the sample gets smaller, the distribution of differences (sampling distribution) becomes less and less normal.
- This difference starts occurring before 30, but until we get that small, it doesn’t make much difference.
- For samples less than 30, the departure from normalcy is so great a special set of curves have been developed to describe these distributions.
- Tables were developed by William Gossett (student)