Data triangulation in the context of qualitative coding and text entropy

Abstract One of the central methodological challenges in qualitative research is ensuring the validity and reliability of coding procedures, especially when analysing complex expert texts. Based on the principle of disciplined subjectivity, triangulation offers a methodological framework that allows for systematic verification of qualitative analyses. This study aims to present a quantitative indicator that, when integrated into qualitative coding, complements traditional triangulation procedures. This study focuses on the application of Shannon’s entropy index and the perplexity derived from it in the evaluation of qualitative coding results. The method was tested on a corpus of expert opinions that were previously processed using qualitative content analysis. During the analysis, we examined the frequency distribution of the codes and calculated the normalised Shannon entropy and perplexity values as indicators of coding balance and diversity. The results indicated that the code groups exhibited high entropy (H_norm ≈ 0.92–0.97) and perplexity values approaching the theoretical maximum (4.4–4.8), indicating that code usage was evenly dispersed rather than concentrated in a few dominant codes. We emphasise that entropy measures the evenness of the code frequency distribution and not the theoretical, semantic, or diagnostic validity of the codes themselves; a balanced distribution is therefore a necessary descriptive property of a well-spread code system, but not sufficient evidence of its validity. Accordingly, entropy is presented here as a complementary, distribution-level descriptor that can be reported alongside—not in place of—established reliability and trustworthiness checks in triangulation analysis. The approach presented herein may enhance the quantitative validation possibilities of qualitative methodology, particularly in research contexts where intersubjective verification alone is insufficient.