Statistical Analysis
Variables of age, years of education and duration of work were dichotomized based on their
distributions in studied subjects. Continuous variables, such as noise level, for different groups
of subjects with normal distribution and equal variance were described using means and standard
deviations, and compared using Analysis of Variance (ANOVA) followed by the least significant
difference (LSD) test for pair-wise comparisons. Continuous variables that did not met
the conditions for ANOVA were described by median and interquartile range (IQR), and compared
using Kruskal-Wallis test followed by Nemenyi test for pair-wise comparisons [19]. Categorical
variables were compared using the Chi-square test or Fisher exact test.
Selected characteristics (including gender, age, years of education, smoking status, etc.),
noise levels, result of audiometric testing and prevalence of hearing loss were compared among
group 1, group 2 and group 3 of the study subjects. Odds ratios (OR) and the corresponding
95% confidence intervals (CIs) for different group as well as potential confounders associated
with hearing loss were estimated using multivariate unconditional Logistic regressions. Because
there was no female worker in group 3 of subjects, we also estimated ORs and 95%CIs by restricting
the regression analysis in male workers only. All analyses were conducted with SAS
version 9.1. All given P values are two-sided