Skip to content
No Vape

A vaping regression table has two reference categories: keep both

Identify the outcome’s base category and a predictor’s reference separately before summarizing a multinomial vaping result.

No Vape · Updated

Disposable vape resting on concrete
Illustrative image · Markus Winkler / Unsplash

Identify the outcome comparison first

An original Oxford Academic table on vaping-harm perceptions presents several response categories against a stated base response, while its packaging rows have their own reference condition. This is a real example of two kinds of reference appearing in one table. We use its labels as a reading example and make no claim about actual product harm.

Multinomial models can address outcomes without a natural order. Keep the categories and the exact base selected by the source. Do not assume that the first visible column, the largest group or the category you personally prefer must be the base. The table or methods should identify it.

Oxford Academic / Nicotine & Tobacco Research: Table 3: Multinomial Logistic Regression Model Predicting Harm Perception Responses

Write a comparison with four labels

Use an invented example with outcome categories A, B and C, where C is the base, and predictor groups X and Y, where Y is the reference. A reported ratio for A in group X concerns the A-versus-C relationship in X relative to that relationship in Y. It does not directly compare A with B.

Write A against C above the result and X against Y beside it. These four invented letters prevent one reference from disappearing when you turn a table into prose. They supply no study data, probabilities or quitting outcome. A different requested comparison needs its own reported analysis or appropriate interpretation, not a guessed subtraction of columns.

Check what the displayed ratio represents

Stata’s multinomial manual explains its base-outcome specification and presentation as relative-risk ratios. The name refers to a ratio of category-probability ratios across predictor values. Read the actual report’s measure and terminology rather than treating the abbreviation as an ordinary binary outcome risk ratio.

Also distinguish a transformed ratio from an untransformed coefficient. A table labeled coefficient, odds ratio or relative-risk ratio requires its stated definition. Do not use a number on one scale as if it were already on another. This explanation does not establish which software the vaping paper used or authorize a fresh analysis of its data.

StataCorp: mlogit: Multinomial (polytomous) logistic regression

Keep group comparisons separate from personal movement

Categories in a model need not represent stages that people travel through in order. A cross-sectional comparison does not show that an adult moved from C to A, even if the category names sound like a progression. Model-derived latent classes are another separate issue: here the first task is identifying a comparison among the specified outcomes.

End with the outcome contrast, predictor contrast, estimate scale and actual population, retaining the reported uncertainty. If either reference remains unclear, leave the conclusion unresolved. This reading exercise does not evaluate a clinical intervention or choose treatment; it helps you avoid changing the question while shortening a technical table into a sentence.

Sources and scope

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

  1. Oxford Academic / Nicotine & Tobacco Research — Table 3: Multinomial Logistic Regression Model Predicting Harm Perception Responses
  2. StataCorp — mlogit: Multinomial (polytomous) logistic regression
  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