Introduction

Visualizing the data may help you find trends worth exploring. Ultimately, however, you will need to perform some statistical tests to see whether these trends may be due to chance or not.

The point of statistical testing is determining which effects are likely to be just random occurrences and which are likely to be true phenomena. If you are predicting the result of a coin toss, you have a 50% chance to be correct every time the coin is tossed. Thus, if you are correct in one toss, you may have an accuracy of 100%, yet this does not prove that you have the ability to predict the result. You would need to be correct multiple times in a row to convince the observers of your capacity. Ultimately, however, no matter how many times in a row you would be correct, you would not prove your ability, you would only do something extremely unlikely.

The tests you will learn here will help you with the most common tasks and comparisons used in linguistics. Similar as in the coin toss examples, they do not prove that your observations are correct or that one variable causes the other, they only give the probability of your results to be due to chance.

Before we move to the tests themselves, a quick note: the tests you will learn here are parametric. That means that in order to use them, some conditions have to be fulfilled. After each test, you will also be shown the non-parametric alternative.
Why should you use parametric tests whenever possible? Because they have better power – they are better at distinguishing real differences from apparent ones.