Compares the means of two sets of data – tests if means are significantly different.
Must meet certains assumptions to be valid:
• The population from which the samples are drawn must have a normal distribution
• The samples are randomly drawn from the population
• The sample have approximately equal variance
The calculated t value is a ratio between the observed difference between the groups to the expected difference.
Two types: 1) t-test for independent or unpaired samples
2) t-test for paired sample (if ‘before’ & ‘after’ with same subject)
If repeated measures design – samples are somehow correlated (eg pretest-post-test design) – samples are not randomly drawn from population – correction can be made in the formula used for standard error of the difference or use a different statistical test.
The purpose of the t-test is to compare two means – it is not appropriate to be used for comparing more than two means. If use multiple t-tests more likely to commit a type 1 error.