# The Role of Statistical Power: A Study of Relationship Between Emotional and Conduct Problems, Sociodemographic Factors, and Smoking Behaviours in Large and Small Samples of Latvian Adolescents

## TL;DR

This study explores how sample size affects the interpretation of factors influencing adolescent smoking in Latvia, emphasizing the importance of considering effect sizes alongside statistical significance.

## Contribution

The study demonstrates how large sample sizes can lead to significant results for weak predictors, advocating for effect size interpretation in public health research.

## Key findings

- Age and conduct problems consistently predict adolescent smoking across sample sizes.
- Large samples yield significant results for weak predictors like family affluence.
- Effect size interpretation is crucial to avoid overstating findings in large datasets.

## Abstract

Background and Objectives: Adolescent smoking is influenced by sociodemographic and psychological factors, including emotional and conduct problems. Understanding how sample size impacts the interpretation of these associations is critical for improving study design and public health interventions. This study examines the relationships between smoking behaviours, sociodemographic factors, and emotional and conduct problems, focusing on how sample size affects statistical significance and effect size interpretation. Materials and Methods: Data from the Latvian Health Behaviour in School-aged Children study was analysed. Chi-square tests and logistic regression were conducted to evaluate associations between smoking behaviours, sociodemographic factors, and emotional and conduct problems. Analyses were performed on both a large general sample and ten smaller generated subsamples to compare the impact of sample size on statistical outcomes. Results: Age and conduct problems emerged as the most consistent predictors of adolescent smoking behaviours across large and small samples, while other predictors, such as family affluence and family structure, showed weaker and less consistent associations. A large sample produced significant results even for weak predictors. Conclusions: This study highlights the importance of integrating effect size interpretation with statistical significance, particularly in large datasets, to avoid overstating findings. By leveraging real-world data, it provides practical recommendations for improving study design and interpretation in behavioural, medical, and public health research, contributing to more effective interventions targeting adolescent smoking.