Answer:
Margin of error - This are the discrepancies that emerge between the results of a sample, and the results of the population as a whole from where the sample was taken from. The smaller the sample, the higher he margin of error.
Polling sample - This is the group of people, extracted from the total population (total people who can vote), that is effectivelly polled. Good pollsters usually use polling samples of around 1,000 people, which gives them a margin of error of around 3%.
Bradley Effect - This effect occurs when a white person is polled about his preferred candidate in a race where there is a white candidate and a minority candidate. In this situation, the white person may say that will vote for the minority candidate, even if he or she will actually vote for the white candidate, in order to avoid in criticism.
Likely voter - as the name implies, this term groups the people who are actually likely to vote, since people who do not vote are very numerous and sometimes they are polled as well, so a distinction is needed.