Asking a Statistical Question

"How tall am I?" has one true answer. Measure yourself once and you're done; there is no variability to explain. "How tall are ninth graders?" is different: every ninth grader gives a different answer, so you have to collect many measurements and describe the spread. That anticipation of variability is what makes a question statistical. It is not about having numbers in the question. "How old am I?" has a number in the answer but still isn't statistical, because there's nothing to investigate beyond one fact.

Turning a vague idea into a statistical question takes three moves. Start with something fuzzy, like 'are teens stressed,' and pin down a measurable variable: maybe self-reported stress level on a 1-to-10 scale, or hours of sleep per night. Then name the population you actually care about: all teens everywhere, or more realistically, students at one school. Finally decide how you'll collect the data: survey, measurement, observation. 'Do students sleep enough' becomes 'How many hours of sleep do ninth graders at this school get on a typical night?' with the variable being hours of sleep, the population being ninth graders at that school, and the plan being a survey or sleep log.

Once a question is sharpened this way, it can do two different jobs. It can describe a single group, like the sleep question above, or it can compare groups, like asking whether students who play sports sleep more than students who don't. That shift from describing to comparing changes what you collect: now you need the sleep data split by group, not just pooled together. This sets up the rest of the unit, where you summarize and compare distributions once the data is in hand.