CosinorOnline: A free web-based platform for cosinor analysis, validated using simulated datasets and demonstrated...

Cosinor analysis remains a fundamental statistical method in chronobiology for quantifying rhythmic patterns. However, many existing tools require specialised programming knowledge or local software installation, creating barriers for clinicians and students. This paper presents the updated CosinorOnline platform, a free browser-based implementation of the established single-component cosinor method designed for accessible rhythmic data analysis. To verify the numerical correctness of the software implementation, CosinorOnline was benchmarked against the reference cosinor2 R package using 2,016 simulated datasets. Mathematical validation was complemented by a cross-platform software benchmark using representative hourly heart-rate profiles derived from continuous radiotelemetric recordings from one control and one prenatally hypoxic rat exposed to endotoxin challenge to verify numerical agreement under realistic physiological conditions. Results demonstrated excellent numerical agreement with established R-based frameworks (Mean Absolute Error $<{10}^{-14}$), consistent with standard floating-point precision. The software also maintained numerical agreement between implementations for the examined non-stationary physiological time-series. The platform provides a benchmarked implementation of the established single-component cosinor analysis, including confidence interval estimation for MESOR and amplitude based on the Student's $t$-distribution. Acrophase is reported as a point estimate without an uncertainty interval. It additionally integrates an ancillary Generalised Lomb-Scargle periodogram as a preliminary exploratory tool for empirical identification of a suitable period ($T$) prior to parametric cosinor analysis, together with interactive polar plot visualisation for phase-amplitude assessment. In summary, CosinorOnline provides an accurate, freely accessible, and reproducible resource for analysing biological time-series data without requiring programming expertise.