A Comparison of Statistical Tests for Likert-Type Data: The Case of Swearwords

Abstract

There has been a long-standing debate about the applicability of different statistical tests to Likert-type data. This work contributes to this debate by comparing the results of three statistics, Analysis of Variance, Kruskal-Wallis H test, and ordinal logistic regression, in the context of self-reported attitude and usage of swearwords. The aims of this comparison are to determine how similar the results of the different statistics are, which of the statistical test are most appropriate for sociopragmatic linguistic investigations, and how the results influence the interpretation possibilities of the same data. An analysis of the results for the different statistics shows some overlap between the three statistics, but that the parametric ANOVA is substantially more conservative in identifying significant relationships between sociodemographic factors and the usage and attitude toward swearwords, while the H test and ordinal logistic regression models are mostly identical, except where the assumptions of the regression model are violated. Based on our results, we argue that the Kruskal-Wallis H test, with the associated post-hoc test, is the most appropriate test for Likert-type data, with sufficient descriptive power to allow for detailed and informed analysis of this data.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 13 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.