Mathematics
5 years ago

Preliminary tests for population variances unreliable under non-normality, study finds.

RePEc: Research Papers in Economics
Preliminary tests of homogeneity- type I error rates under non-normality
Tanweer Ul Islam

Paper Summary

The study looked at tests used to check if groups have similar variability, even when data is not normally distributed. The Goldfeld-Quandt and Levene tests can give inaccurate results under non-normal conditions, with errors as high as 88% and 48% respectively. However, a modified version of the Levene test (BF-test) still works well, except in cases with big outliers.

Related papers

Theme background image
Mathematics
8 years ago

New tests detect non-normality in data, improving accuracy of statistical analysis.

Multicollinearity
Regularization Methods
Statistical Control
1 view
Theme background image
Agricultural and Biological Sciences
7 years ago

F test unreliable for comparing variances, leading to inaccurate results

Pollinator Decline
Population Genetics
Genetic Adaptation
Theme background image
Social Sciences
8 years ago

SAS PROC TTEST®: Unveiling Truth Behind Population Mean Differences and Variance.

Social Sciences Methods
Plasma Applications
Multicollinearity
1 view