Skip to content
No Vape

What does a VIF check tell you in vaping research?

Read a predictor-collinearity diagnostic without treating a low value as proof of independent causes, complete adjustment or a clinically valid model.

No Vape · Updated

Open book on a cafe table
Illustrative image · Lucille Emi Oh / Unsplash

Identify the diagnostic rather than the health claim

A methods paragraph may say the authors checked VIF and found little collinearity. Before repeating that as proof that several factors act independently, locate the actual analysis and its predictors. In this context, independent variable is a conventional model role. It does not mean the variable is causally independent of everything else.

Nakju and colleagues’ 2026 Bangkok school-student report supplies a VIF statement in the Table 4 footnote. It concerns predictors in an analysis of intention to use e-cigarettes. That is a real example of the diagnostic appearing in vaping research, not a basis for assigning a US adult an intention score, diagnosing dependence or inferring a quitting outcome.

Saroj Nakju and colleagues / Tobacco Induced Diseases: Prevalence and factors associated with intention to use e-cigarettes among lower secondary school students in Bangkok

Ask what overlap the check addresses

NIST describes multicollinearity as substantial linear interdependence among columns of a regression design matrix. Such overlap can make coefficient estimates unstable. A VIF is one diagnostic for that problem; it is different from asking whether one exposure is associated with the outcome or whether important confounders were measured.

Consider invented predictors that describe the same appointment interval once in minutes and once in hours. They are two columns, but one is exactly determined by the other. A fictional model cannot treat that duplication as two separate pieces of independent information simply because the headings differ. This is a mathematical illustration, not a recommended vaping model or an actual study finding.

NIST Dataplot: Variance Inflation Factors

Retain the checked set and avoid a universal pass mark

Write the predictor list or the stated model reference next to the diagnostic. A check for one specification does not automatically cover a later model with additional interactions or changed coding. Note whether the report supplies individual values, a range or only a summary, and retain any unclear detail as a question.

You do not need to choose a preferred VIF cutoff to read the claim accurately. Nor should you remove a variable from your own records because a methods paragraph mentions redundancy. Decisions about a research model require its question and assumptions. This guide explains the purpose of the stated check without auditing the named report or endorsing every part of its analysis.

Separate numerical stability from causal adequacy

Even a predictor set with little detected linear overlap can omit a relevant influence, use imprecise measurements or answer an association question. Conversely, a collinearity concern does not by itself establish that the measured phenomenon is imaginary. Keep the uncertainty about coefficients separate from claims about why people vape.

End with a bounded reading card: actual model, checked predictors, reported diagnostic and what that diagnostic does not test. If a headline moves from low VIF to “these are proven independent causes,” ask for the missing causal evidence. Personal support choices and treatment decisions need their own appropriate information; a design-matrix check is not a health recommendation or a certification of an app.

Sources and scope

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

  1. Saroj Nakju and colleagues / Tobacco Induced Diseases — Prevalence and factors associated with intention to use e-cigarettes among lower secondary school students in Bangkok
  2. NIST Dataplot — Variance Inflation Factors
  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