I am preparing some materials for an introduction to sample size calculation for life sciences graduate students with no or very limited previous statistical experience. Right at the start I will be defining the basic distributions (normal, binomial, Poisson and possibly exponential). Mostly, this is trivial (“Poission distribution describes probability of a given number of events occurring in a specified interval of time or space”, “binomial distribution describes number of results of a particular type when a process with two possible outcomes is samped a certain number of times”, etc) but I’ve noticed that I normally describe the general shape of the distribution and give some examples of the types of real-world processes that result in a normal distribution (heights of a group, that sort of thing) rather than define it in terms of the general class of processes that the distribution describes. Does anyone have any experience of what works best (in terms of student comprehension) in this situation?
The most obvious way I can see is to do it in terms of the central limit theorem – something like ‘the normal distribution describes the distribution of the average of groups of measurements of most random variables. The distribution of these averages approaches the normal distribution more closely as the number of groups increases.’ Any thoughts on this as a working definition? I will then give some real-world examples, but I want a stand-alone definition first. I don’t think I need to go into things like finite variance given the audience (and similarly I think they will interpret ‘average’ as ‘arithmetic mean’ by default) – I want to keep the language as accessible as possible.
(I'm aware that real-world data of this type is not necessarily normally-distributed in practice.)