A vaping-related count model uses an offset: find the rate underneath
Check the denominator built into a Poisson-model result instead of interpreting a per-population comparison as a raw count difference.
No Vape · Updated

Find what the model counts and what scales it
The Ontario vaping-retailer study describes a Poisson model whose outcome is a retailer count around a defined neighborhood center, with the population aged fifteen or older used as an offset. The paper explains that its results represent retailers per capita. That is an actual example of a vaping-related count and a separate scaling denominator.
The historical Canadian setting and geographic units remain part of that example. It neither estimates a US reader’s vaping nor records individual quitting events. Locate these two ingredients in your own source: the counted quantity and the denominator attached through the offset. The word exposure in software terminology need not mean chemical exposure.
Use a small count-versus-rate exercise
Imagine two invented observation units. One records two ordinary events across ten units of observation; the other records four across twenty. Their counts differ, while their simple event-per-observation ratios match. This arithmetic does not fit a model, but shows why a comparison of raw totals can answer a different question from a rate comparison.
The example has no real people, clinical events or recommended monitoring schedule. Keep the denominator’s meaning explicit: time, population and sampled area are not interchangeable. Do not convert a population offset into days of follow-up or assume that two equal ratios prove the underlying processes are identical.
Distinguish an offset from a regular predictor
Stata’s official explanation writes expected count as exposure multiplied by the modeled rate. It also explains that an unspecified exposure is treated as one in that presentation. This supports reading the scaling role, not choosing a statistical model for your own observations or asserting that the Ontario researchers used that software.
Ask where the report documents the offset and whether a displayed result is a count, rate, rate ratio or coefficient on another scale. A table heading and methods note should agree. Do not describe a rate ratio as a percentage-point difference or assume an adjusted label answers which denominator was used.
Keep the remaining model questions separate
Accounting for a denominator does not establish representative sampling, absence of confounding or an adequate variance model. Those remain distinct questions. A paper may use a negative-binomial model, random effects or sensitivity checks alongside an offset; none changes the need to identify the actual target before explaining a reported number.
Finish with the outcome count, scaling variable, geographic or time unit and reported estimate type. If the denominator is unavailable, retain that uncertainty rather than reconstructing it from a rounded result. This is a reading map, not a model-fitting tutorial or clinical appraisal. Your personal progress record does not become a validated rate analysis because you divide two numbers.
Sources and scope
Sources inform the guidance. The worksheets and examples are No Vape editorial suggestions, not tested treatments.
Written with AI assistance and checked against the cited sources. No clinical review is claimed. This guide covers practical planning, not individual medical treatment. How we research and write