We start talking about probability distribution in chapter 2.2.3 Normal Distributions & Normality. So you already know that:
- When you don’t have all the process data, you need to work on a representative sample;
- Plotting this sample with descriptive statistics (like a graph), help you to find the correct type of distributions;
- With the correct type of distribution, you can assess the probability of an event in a process.
Finally, knowing the probability of an event in a process can help you to reply to questions like:
- How many non-conforming units can I have in a process?
- What’s the number of non-conformity can I have in a process? (maybe a unit with more than one error)
- How many failures can I have in a trial before the first success?
In this case, you are basing your response on only a tiny sample instead of counting the entire population, which in the same case, it’s impossible or too expensive.
The main kind of distribution are:
- Discrete distribution: when you have a variable that assumes only a countable value. For example, the face of a coin are two and are countable;
- Continuos distribution: when you have a variable that assumes a non-countable value. So the value that it can be assumed is infinite in a range. For example, the temperature in a room can be considered to take an endless number of values.