Applying the articles to the dataset responsibly

This lesson shows how to use ideas from Beard et al. 2017 and Fa-Binefa et al. to ask sensible exploratory questions about the workbook Copy of Smoking Data 2(1). The key principle is to match each research idea to the variables that are actually present in the dataset, and to avoid claiming that the workbook contains article-level measures when it does not.

1. Start with what the workbook actually contains

According to Copy of Smoking Data 2(1), the available variables are:

• Age – scale variable
• Alcohol – nominal variable, with categories for alcohol consumption
• AgeOnset – nominal variable for age of onset of smoking
• Gender – nominal variable
• Smoking – scale variable for smoking behaviour

The workbook also gives value labels:

• Alcohol: 0 = none to low levels; 2 = moderate to high levels
• AgeOnset: 1 = 0 to 30 years age; 2 = 31 and older
• Gender: 1 = male; 2 = female

This means the dataset is suitable for descriptive analysis and simple group comparisons, because it includes one continuous smoking measure and several grouping variables.

2. Use the articles to frame exploratory questions

Beard et al. 2017 examined links between smoking and alcohol-related behaviour at the population level. Their study focused on associations between smoking prevalence and high-risk drinking prevalence, as well as motivation and attempts to change these behaviours. This supports asking a simpler workbook-based question such as:

• Do people in different alcohol-consumption categories show different average smoking behaviour?

Fa-Binefa et al. focused on smoking onset age as an exposure indicator and examined how earlier smoking onset related to later health risks and mortality. That article supports asking workbook-based questions such as:

• Does smoking behaviour differ between people with earlier versus later smoking-onset groups?
• Does age of onset appear to vary alongside age or gender in this sample?

Because the workbook also includes age and gender, you can extend these into basic subgroup questions:

• Does smoking behaviour differ by gender?
• Does smoking behaviour appear to increase or decrease with age?
• Within alcohol groups, are there gender differences in smoking behaviour?
• Within age-of-onset groups, are there age or gender differences in smoking behaviour?

3. Variables that can support simple descriptive analyses

Using Copy of Smoking Data 2(1), the following variables can be used for straightforward summaries:

• Age: mean, median, range, and distribution summaries
• Smoking: mean, median, range, and spread of smoking behaviour scores
• Alcohol: counts and percentages in each alcohol category
• AgeOnset: counts and percentages in each onset category
• Gender: counts and percentages by sex category

These summaries are appropriate because they stay close to the measurement level shown in the workbook documentation.

4. Variables that can support simple group-comparison analyses

The workbook is especially well suited to comparing the continuous Smoking variable across categories. For example:

• Smoking by Alcohol group: compare average smoking behaviour for none/low versus moderate/high alcohol groups
• Smoking by AgeOnset group: compare average smoking behaviour for onset at 0–30 years versus 31+ years
• Smoking by Gender: compare average smoking behaviour for males and females

You can also explore simple relationships involving age:

• Age and Smoking: examine whether older participants tend to have higher or lower smoking behaviour scores
• Age by Alcohol group: compare average age across alcohol categories
• Age by Gender: compare age distributions across gender groups

Cross-tabulations are also possible among the categorical variables:

• Alcohol by Gender
• Alcohol by AgeOnset
• Gender by AgeOnset

These are reasonable exploratory analyses because they rely only on variables directly available in Copy of Smoking Data 2(1).

5. How the articles should and should not be applied

The articles are useful for conceptual framing, but the workbook does not reproduce their full study designs.

From Beard et al. 2017, you can borrow the broad idea that smoking and alcohol-related behaviour may be associated. However, their study used monthly representative survey data and time-series ARIMAX modelling to assess changes in prevalence over time. The workbook does not include monthly waves, repeated time points, or prevalence series. So it would be inappropriate to describe your workbook analysis as a time-series test of population trends.

From Fa-Binefa et al., you can borrow the idea that age at smoking onset matters. But their study linked smoking onset age to cardiovascular disease events, cardiovascular mortality, and all-cause mortality using cohort follow-up and Cox regression. The workbook contains no disease outcomes, no follow-up period, and no mortality data. So you cannot test the article's main health-outcome claims with this dataset.

6. Important article-level constructs not directly available in the workbook

The workbook lacks several constructs central to the two articles.

Not directly available from Beard et al. 2017:

• Smoking prevalence at the population level
• High-risk drinking defined by AUDIT score
• Motivation to quit smoking
• Attempts to stop smoking
• Motivation to reduce alcohol consumption
• Attempts to reduce alcohol consumption
• Monthly repeated observations for time-series analysis
• Policy exposure or interrupted time-series variables

Not directly available from Fa-Binefa et al.:

• Exact smoking-onset age as a continuous measure
• Smoking status categories such as current, former, and never smoker
• Pack-years or cumulative tobacco exposure
• Years since quitting smoking
• Cardiovascular disease events
• Cardiovascular mortality
• All-cause mortality
• Clinical covariates such as cholesterol, glucose, and hypertension treatment
• Longitudinal follow-up time for survival analysis

This is especially important for responsible interpretation: the presence of variables called Alcohol, AgeOnset, and Smoking does not mean the workbook contains the same constructs used in the published studies.

7. Responsible wording for exploratory conclusions

When writing up findings, keep the language modest. Because this workbook supports only simple exploratory analysis, better wording would be:

• “In this sample, smoking behaviour scores differed across alcohol-consumption groups.”
• “In this dataset, the age-of-onset category was associated with different smoking behaviour levels.”
• “These patterns are consistent with questions raised in Beard et al. 2017 and Fa-Binefa et al., but the workbook does not include the full measures or design needed to test those studies directly.”

Avoid stronger claims such as:

• that alcohol causes smoking changes
• that early onset causes cardiovascular disease
• that the results estimate population prevalence trends
• that the workbook measures motivation, quitting attempts, mortality, or clinical risk

8. A practical set of exploratory questions for this workbook

A responsible question set, grounded in the articles but limited to the available variables in Copy of Smoking Data 2(1), would be:

• What is the distribution of smoking behaviour in the sample?
• How are participants distributed across alcohol-consumption categories?
• How are participants distributed across smoking-onset categories?
• Is smoking behaviour different between alcohol groups?
• Is smoking behaviour different between age-of-onset groups?
• Is smoking behaviour different between males and females?
• Is age related to smoking behaviour?
• Do alcohol group and age-of-onset group together help describe variation in smoking behaviour?

9. Key takeaway

The most responsible way to apply Beard et al. 2017 and Fa-Binefa et al. is to use them as theoretical guides for asking questions, not as templates for claiming the same analyses or conclusions. In Copy of Smoking Data 2(1), the variables Age, Alcohol, AgeOnset, Gender, and Smoking can support descriptive summaries and basic group comparisons. But important article-level constructs—such as high-risk drinking scores, motivation, quit attempts, disease outcomes, mortality, and longitudinal follow-up—are not directly available in the workbook.