Health surveys are one of the most prominent sources of information in medical sciences. Analysis of a health survey not only gives information about the inherent features of the population under study but also sheds light on the interactions between different health factors. Even though there has been a multitude of methodologies proposed for survey analysis, many of them are biased towards the previous hypotheses, the fact which may lead to spurious results by neglecting the whole structure of relations among the factors. Here, we proposed our novel workflow on survey analysis. By gathering the previously less-utilized graphical models on survey analysis, newly-proposed methods, and commonly-used tools of prediction and hypothesis testing our workflow provides the medical community with an end-to-end pipeline for comprehensive statistical analysis of health surveys. To demonstrate the functionality of our workflow, we validate it on the Yazd Health Study (YaHS) dataset, in which a questionnaire of 300 questions from 10000 participants associated with 40 laboratory measurements from a subset of 4010 individuals is acquired from the residents of the Yazd greater area of Iran. Many of our findings are aligned with previous medical knowledge. Most of our novel findings are surprisingly observed in the clinics, each of which is potential to raise a promising hypothesis.

A Novel Workflow for Health Survey Analysis: Results on the Yazd Health Study (YaHS)