COMPARATIVE STATISTICAL INFERENCE OF PM2.5 LEVELS ACROSS INDIAN CITIES : A BOOTSTRAP vs CLASSICAL APPROACH

Abstract

Air pollution remains a pressing environmental and public health challenge in India, with fne particulate matter (PM2.5) posing severe respiratory and cardiovascular risks. This study conducts a comparative statistical inference analysis of daily PM2.5 concentrations for Delhi and Mumbai, based on 2024 data sourced from the Central Pollution Control Board (CPCB). Two estimation approaches are applied: the classical parametric t-based confidence interval method, which assumes normality, and the non-parametric bootstrap approach, which relies on re-sampling without distributional assumptions. The analysis reveals that while Delhi consistently exhibits substantially higher PM2.5 levels than Mumbai, the estimated means and confidence intervals from both methods are closely aligned, indicating that the parametric method’s assumptions are reasonably met in this dataset. The findings underscore the utility of bootstrap methods in validating classical inference, particularly in environmental data analysis, and provide robust evidence for policy-oriented air quality interventions.

Keywords: BOOTSTRAP APPROACH, CLASSICAL APPROACH,confidence intervals , Exploratory Data Analysis