Statistical Assumptions Matter in Data Analysis for Texas Ozone Nonattainment Sites

Strategies for control of ozone aim at regulation of its chemical precursors, non-methane organic compounds (NMOC) and nitrogen oxides (NOx). It is therefore important to analyze how these precursors vary temporally and geographically. This study finds significant and important differences among four Texas ozone nonattainment sites, Dallas, Ft. Worth, El Paso, and Houston, for 1984, 1985, and 1986 for NMOC, NOx, and their ratio NNR. These differences were detected through nonparametric analysis of variance and Student-New-man-Keuls’s test for multiple comparisons on rank-transformed data. A noteworthy feature of the data analysis is its attention to the assumptions underlying the statistical methods. Classical models based on normal or lognormal theory had to be abandoned for lack of realism. It is demonstrated how alternative models may be applied to yield appropriate, rather than inappropriate, conclusions.