How do i know if my data is normally distributed?

Question

So apparently you should use t-test for paired observations, when n is small but normal distribution can be assumed. How do i know when it can be assumed? And how 'small' is small? n<50?  if so, is it 50 pairs or 50 total observations?

Answer 1

I call this a Paried t test.

If you subtract the values from each other within each pair, you get a new column of data (paired differences). Those differences need to fit a normal distribution (or not be too far from the distribution shape).

You don't need 50 data points (although more is better). You need at least 20 data points to get a rough ideaof the shape of the distribution (using a histogram), or you can conduct an Anderson-Darling normality hypothesis test. However, small samples may not have enough power to reject normality (default assumption is that data is normal). If you can get 30+ data points, that would be better.