In stats, how would u know if a t-test was appropriate?

Answer 1

A t-test is appropriate if:

the data is interval or ratio

if you are comparing only 2 "sets" of compatible/related data

So a t-test would be appropriate to compare performance on the same test (that's the compatible or related data) between 2 classrooms (independent samples t-test).

Or you could use a t-test to compare performance on a pre-test and a posttest taken by the same group of people (paired samples t-test).

Answer 2

You use a t-test when the population is normal but you don't know the sd and need to estimate it with your sample.

You'll notice that as your sample gets larger then the estimate of sd becomes better and the t-test get closer to a normal. Look at your t tables and notice that for df->infinity the values approach the normal values.

Answer 3

If n<30, because the sampling distribution of the sample mean wouldn't be approximately normally distributed as guaranteed by the Central Limit Theorem, when n>/=30.

Answer 4

in order to perform a T test you must know if the sample or population meets parametric criteria, such as random selection, independent versus dependant variables, and they cannot be represented by nominal or ordinal scales. The frequency must also follow a normal distribution.