The Importance of the Normality Assumption in Large Public Health Data Sets

Важность предположения о нормальности в больших наборах данных общественного здравоохранения
Thomas Lumley, Scott S. Emerson, Paula K. Diehr, Lu Chen
2002-05-01

Normality assumptionWilcoxon rank sum testlinear regressionordinal logistic regressiont-test
It is widely but incorrectly believed that the t-test and linear regression are valid only for Normally distributed outcomes. The t-test and linear regression compare the mean of an outcome variable for different subjects. While these are valid even in very small samples if the outcome variable is Normally distributed, their major usefulness comes from the fact that in large samples they are valid for any distribution. We demonstrate this validity by simulation in extremely non-Normal data. We discuss situations in which in other methods such as the Wilcoxon rank sum test and ordinal logistic regression (proportional odds model) have been recommended, and conclude that the t-test and linear regression often provide a convenient and practical alternative. The major limitation on the t-test and linear regression for inference about associations is not a distributional one, but whether detecting and estimating a difference in the mean of the outcome answers the scientific question at hand.
1
For small samples, t-tests and linear regression are valid when the outcome is Normally distributed.
2
Simulation demonstrates that t-tests and linear regression retain validity in large, highly non-Normal public health datasets.
3
The main limitation is scientific relevance—whether differences in outcome means adequately address the association question—not distributional non-Normality.
4
The t-test and linear regression are valid in large samples regardless of the outcome distribution, including extremely non-Normal data.
5
Wilcoxon rank-sum tests and ordinal logistic regression are not always necessary; t-tests and linear regression often offer practical alternatives.

t-test and linear regression applied to large public health data sets with non-Normal outcome distributions

Validity and practical applicability for comparing and estimating mean outcomes in large samples regardless of outcome distribution

Publication Details
Publication Date
2002-05-01
Journal
Publisher
ISSN
Cited by
1985
Access Type
Author Information
Authors
Thomas Lumley
Scott S. Emerson
Paula K. Diehr
Lu Chen
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%