Consistency errors in p-values reported in Spanish psychology journals

José Manuel Caperos Montalbán, Antonio Pardo

Psicothema · 2013 · 25 citations · 18 references

DOIFull text

Open access

Concepts

Abstract

BACKGROUND: Recent reviews have drawn attention to frequent consistency errors when reporting statistical results. METHOD: We have reviewed the statistical results reported in 186 articles published in four Spanish psychology journals. Of these articles, 102 contained at least one of the statistics selected for our study: Fisher-F , Student-t and Pearson-c 2 . RESULTS: Out of the 1,212 complete statistics reviewed, 12.2% presented a consistency error, meaning that the reported p-value did not correspond to the reported value of the statistic and its degrees of freedom. In 2.3% of the cases, the correct calculation would have led to a different conclusion than the reported one. In terms of articles, 48% included at least one consistency error, and 17.6% would have to change at least one conclusion. In meta-analytical terms, with a focus on effect size, consistency errors can be considered substantial in 9.5% of the cases. CONCLUSION: These results imply a need to improve the quality and precision with which statistical results are reported in Spanish psychology journals.

References

18