<abstract xmlns="http://www.w3.org/1999/xhtml"><p>During the last two decades, an increase of tobacco product reporting requirements from regulators was observed, such as Europe, Canada or USA.</p><p>However, the capacity to compare and discriminate accurately two products is impacted by the number of constituents used for the comparison. Indeed, performing a large number of simultaneous independent hypothesis tests increases the probability of rejection of the null hypothesis when it should not be rejected. This leads to virtually guarantee the presence of type I errors among the findings. Correction methods have been developed to overcome this issue like the Bonferroni or Benjamini &amp; Hochberg ones. The performance of these methods was assessed by comparing identical tobacco products with different sizes of data sets. Results showed that multiple comparisons lead to erroneous conclusions if the risk of type I error is not corrected. Unfortunately, reducing the type I error impacts the statistical power of the tests. Consequently, strategies for dealing with multiplicity of data should provide a reasonable balance between testing requirement and statistical power of differentiation. Multiple testing for product comparison is less of a problem if studies restrict to the most relevant parameters for comparison.</p></abstract>

During the last two decades, an increase of tobacco product reporting requirements from regulators was observed, such as Europe, Canada or USA.However, the capacity to compare and discriminate accurately two products is impacted by the number of constituents used for the comparison. Indeed, performing a large number of simultaneous independent hypothesis tests increases the probability of rejection of the null hypothesis when it should not be rejected. This leads to virtually guarantee the presence of type I errors among the findings. Correction methods have been developed to overcome this issue like the Bonferroni or Benjamini & Hochberg ones. The performance of these methods was assessed by comparing identical tobacco products with different sizes of data sets. Results showed that multiple comparisons lead to erroneous conclusions if the risk of type I error is not corrected. Unfortunately, reducing the type I error impacts the statistical power of the tests. Consequently, strategies for dealing with multiplicity of data should provide a reasonable balance between testing requirement and statistical power of differentiation. Multiple testing for product comparison is less of a problem if studies restrict to the most relevant parameters for comparison.

Analysis of the Effect of Multiple Testing in Assessing Tobacco Product Differences

Contributions to Tobacco & Nicotine Research

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

During the last two decades, an increase of tobacco product reporting requirements from regulators was observed, such as Europe, Canada or USA.However, the...

Analysis of the Effect of Multiple Testing in Assessing Tobacco Product Differences

Publié en ligne: 22 mai 2017

Pages: 78 - 85

Reçu: 22 nov. 2016

Accepté: 30 mars 2017

DOI: https://doi.org/10.1515/cttr-2017-0008

Mots clés
Tobacco product, mainstream cigarette smoke, smoke toxicant emissions, statistical comparison, measurement variability

© 2017 Beiträge zur Tabakforschung International

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Analysis of the Effect of Multiple Testing in Assessing Tobacco Product Differences

Publié en ligne: 22 mai 2017

Pages: 78 - 85

Reçu: 22 nov. 2016

Accepté: 30 mars 2017

DOI: https://doi.org/10.1515/cttr-2017-0008

Mots clésTobacco product, mainstream cigarette smoke, smoke toxicant emissions, statistical comparison, measurement variability

© 2017 Beiträge zur Tabakforschung International

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Mots clés
Tobacco product, mainstream cigarette smoke, smoke toxicant emissions, statistical comparison, measurement variability