Statistical Evaluation of Glucose Level Distributions among Women: Implications for Thyroid Function Analysis

Abstract

Background: The current study takes up issues of glucose levels in the background of thyroid health among women aged 15-49, with 188 and 365 samples from the NFHS-5 dataset.

Objectives: To study the relationship between glucose levels and thyroid health among women. Whether or not the glucose levels depart from a symmetric distribution. Abnormalities in thyroid function and related metabolic processes shall be investigated. 

Methods: Data were sourced from the NFHS-5 survey. The parameters included the levels of glucose and indicators of thyroid health. A proper study of these data was conducted by fitting 25 distributions, which would indicate the most appropriate model for glucose levels. There are three primary statistical tests that can be employed to evaluate the quality of each distribution’s fit: the Kolmogorov-Smirnov one-sample test, the Anderson-Darling test, and the chi-square test. Results: The four-parameter generalised gamma distribution was ranked as the best fitting model for the glucose dataset, with the beta distribution ranked second. Using all 365 samples, the beta distribution was ranked highest; with a reduced sample size of 188, the best ranking distribution was the gamma distribution.

Novelty: Its novelty comes from a way of understanding the relationship between glucose levels and thyroid disorders in women aged 15-49 years. According to that study, glucose levels were found not symmetrically standardised, hence the potential abnormalities of thyroid function and its associated metabolic process. This has not been extensively documented in the literature and thus presents a novel contribution to both the fields of endocrinology and biostatistics.