Match the Model to the Process

The concept

A probability distribution describes possible outcomes and their probabilities or relative density under a model. Model choice should reflect the process and the question. Counts, waiting times, and measurements are not interchangeable. A familiar bell curve can be a useful approximation in some settings, but familiarity alone does not justify its use.

Worked example

A fictional ticket office counts arrivals in equal ten-minute intervals. Quiet weekday periods and a sold-out event produce very different patterns. Combining both into one distribution gives an average that fits neither setting well. The analyst separates operating conditions, inspects the observed frequencies, and compares proposed models with data from similar periods. The purpose is to predict staffing demand, so rare busy intervals deserve particular attention.

Limits and judgment

All models simplify. A good fit in the middle does not guarantee accurate tail probabilities, and limited observations cannot rule out unusual events. State the conditions under which the model is intended to apply. Check sensitivity to plausible alternatives before using a small calculated probability to dismiss an operational risk that the data barely measure.

RETRIEVE. REASON. REMEMBER.

Knowledge check.

Choose one answer for each question, then check your reasoning.