INFERENCE TESTING
The first concept of inference testing is the test group itself. If you’re conducting an inference test, the sample you use must be representative of population. For instance, if you are trying to prove that 80% of young people (let’s say 25 and under) own cell phones, the sample you take must be of that target age group or else your results will not be valuable. Furthermore, you must randomize. This is a key concept in statistics, as randomization reduces potential bias.
Inference testing has two main branches: estimation and hypothesis testing.
ESTIMATION
Estimation is when a statistician guesses a value of an unknown population parameter and conducts confidence interval to capture the unknown value. A confidence interval is a range that covers where the statistician believes the value could fall in with a degree of certainty, while accounting for natural variation. This type of inference testing is used in voting polls.
HYPOTHESIS TESTING
Hypothesis testing is when a statistician starts with a belief, makes a hypothesis, and collects sample data to lead them to retain or reject their hypothesis. A statistician will never accept a hypothesis due to the fact that statistics are only educated predictions, not facts. Think of it like a court; guilt doesn’t prove innocence. In hypothesis testing, a statistician solves for a “P value”, or the likelihood that the sample results occur by chance. If that P Value is small, this is an indication that there is significant findings, and the initial hypothesis should be rejected.
There is plenty of room for error in inference testing. First, half of the battle is figuring out what type of mechanical test is best to help you answer your question. If you use the wrong test, the results could be futile in solving that curiosity. Also, a lack of randomization or a representative sample could skew affect your results.
Inference testing is an integral part of statistics because is used to say something about a population based on sample data. The information to be gained from inference testing allows statisticians to gain calculated evidence of their speculations.
