Skip to content
No Vape

Zero-inflated and hurdle vaping models do not use zeros in the same way

Identify the zero-related and count components before turning a vaping-model result into a claim about nonuse or quitting.

No Vape · Updated

Books arranged on a shelf
Illustrative image · Pramod Tiwari / Unsplash

Separate the recorded answer from the model’s components

A zero needs the survey’s actual meaning and time window. It could describe no reported use during a specified period rather than never having used a vape. A blank response is another state. Keep that observation separate from the additional structure a statistical model uses to explain a distribution of answers.

An original adolescent vaping paper used zero-inflated models for nicotine and marijuana frequency measures. Its responses were coded frequency categories rather than exact puff counts. We use the documented method only: no peer-influence result, adolescent use estimate or recommendation is being applied to an adult’s quitting plan.

Skinner and colleagues / Addictive Behaviors: My friends made me do it: Peer influences and different types of vaping in adolescence

Locate where zeros can arise

The original countreg documentation describes zero-inflated models as a mixture of a point mass at zero and a count distribution. The count distribution can itself produce a zero. Consequently, the extra-zero component is not simply a list of everyone whose observed response was zero. Its meaning depends on the model and the report’s assumptions.

The same documentation describes hurdle models differently: one component represents zero versus positive outcomes, and the count component is restricted to positive values. Keep those components separate when reading a coefficient table. A result for the positive component is not automatically a result for all respondents or the probability of reporting any use.

Achim Zeileis and Christian Kleiber / countreg original documentation: Zero-augmented models

Build a component map before repeating a headline

Use four fields: recorded outcome and coding; zero-related component; count or frequency component; exact result the summary refers to. Copy the authors’ labels, including which outcome their binary component models. A heading about no use and one about any use point in opposite directions; do not silently swap them.

For a fictional example, a table has separate zero-related and positive-frequency results, but a caption calls both a change in quitting. Your map leaves two different questions visible rather than choosing the more encouraging interpretation. No numbers or real study findings are supplied by this illustration, and it does not establish that a particular published model was fitted correctly.

Keep model membership and personal history apart

A modeled component does not establish that a named respondent will never vape, that a zero was misreported or that an observed absence should be deleted. Do not relabel your own diary as structural zeros to fit a research category. Retain what you actually observed, its coverage and the uncertainty that remains.

This reading exercise does not choose between models, estimate coefficients or translate them into personal odds. The relevant endpoint is identifying which component supports the published statement and which question it leaves open. Consequential interpretations require the full report and appropriate expertise; a practical next action need not wait for a statistical classification of you.

Sources and scope

Sources inform the guidance. The worksheets and examples are No Vape editorial suggestions, not tested treatments.

  1. Skinner and colleagues / Addictive Behaviors — My friends made me do it: Peer influences and different types of vaping in adolescence
  2. Achim Zeileis and Christian Kleiber / countreg original documentation — Zero-augmented models
  3. No Vape — No Vape — app features and store links

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