Skip to content
No Vape

Read a lasso-selected vaping predictor without calling it a cause

Separate coefficient shrinkage and selection from a causal ranking, a personal dependence score or proof that an omitted factor never matters.

No Vape · Updated

Stack of books on a table
Illustrative image · Laura Ohlman / Unsplash

Find what the model was asked to predict

A headline listing the strongest vaping predictors may conceal several choices: which outcome, which measurement occasion, which people and which candidate inputs. Start with those definitions. A model for a questionnaire score is not automatically a model for stopping vaping, and a study population is not every person who uses an app.

Singh and colleagues’ 2023 Canadian study used separate lasso models for baseline daily and nondaily vapers to predict a later vaping-dependence score. This supplies a concrete published use of the method. The study is not a No Vape feature, an adult treatment rule or evidence that a reader should calculate a private score from the published coefficients.

Inderpreet Singh and colleagues / Healthcare: A Machine Learning Approach Reveals Distinct Predictors of Vaping Dependence for Adolescent Daily and Non-Daily Vapers in the COVID-19 Era

Understand why some coefficients can disappear

MathWorks explains lasso as fitting with a coefficient-size penalty. Increasing the penalty can set coefficients to zero, producing a smaller model. Its official explanation distinguishes that selection behavior from ridge regression and describes elastic net as a related approach. These are model-fitting properties, not judgments about the personal importance of a life circumstance.

For an original fictional example, suppose two candidate columns carry very similar schedule information. A penalized fit retains one and drops the other. That does not mean the discarded description is unrelated to real life, or that changing the retained one would alter someone’s outcome. Selection depends on the data, candidate set and fitting choices, rather than revealing a universal list of causes.

MathWorks: Lasso and Elastic Net

Keep tuning separate from a final test

The vaping study describes selecting its penalty through cross-validation and then evaluating predictive performance on a testing set. Retain which role each dataset played. A tuning process searches among fits; that is not the same task as demonstrating performance on genuinely held-out observations or showing benefit from using a prediction system.

Your reading note can name the target, candidate inputs, stated penalty-selection method and evaluation population. Leave a missing detail unresolved rather than assuming the word machine learning guarantees a separate test. A performance metric and a selected coefficient also answer different questions; neither one alone describes how an individual should change a quit plan.

Inderpreet Singh and colleagues / Healthcare: A Machine Learning Approach Reveals Distinct Predictors of Vaping Dependence for Adolescent Daily and Non-Daily Vapers in the COVID-19 Era

Avoid both a causal ranking and a negative verdict

Do not restate retained variables as the things that make every person vape. Likewise, a zeroed coefficient is not proof that a factor never matters clinically or socially. An observational prediction analysis needs separate causal evidence before it can support an intervention claim, even if the fit predicts its stated outcome well.

The useful output is a model-specific description of what was selected under the documented process, with personal conclusions left outside it. If a health decision depends on the paper, seek qualified interpretation of the complete study. You can prepare your own practical observation without uploading private records to an unfamiliar model demonstration or treating a research-selected input as a treatment instruction.

Sources and scope

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

  1. Inderpreet Singh and colleagues / Healthcare — A Machine Learning Approach Reveals Distinct Predictors of Vaping Dependence for Adolescent Daily and Non-Daily Vapers in the COVID-19 Era
  2. MathWorks — Lasso and Elastic Net
  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