Date of Award


Degree Name

Master of Science


Mathematical Sciences


Joan G. Staniswalis


El Paso, Texas is known as one of the dust hotspots in North America. We explore the effect of dust and low wind events on asthma admissions in El Paso, Texas between 2000 and 2005. Conditional logistic regression with a case-crossover design was used to estimate the probability of hospitalization after dust and low wind events while controlling for pollutants with hourly monitor measurements, and weather. The historical functional linear model is used to incorporate the hourly pollutant measures into the regression model with a continuous lag, as an alternative to a distributed lag model based on daily averages. The great advantage of the historical functional linear model was demonstrated, namely, the pollutant past exposure was included without having to choose a (short-term) lag. The nonparametric functional linear model in the conditional logistic regression framework is fit by first preprocessing the data, then applying the COXPH function in the R-package for survival analysis. Based on the theoretical relationship between P-splines and ridge regression we modified COXPH to allow for a penalty in the nonparametric functional linear model, which was a very time saving approach for us. We proposed that the ridge trace be used to guide the choice of the smoothing parameter in nonparametric functional linear model. The results obtained from the simulated examples suggest that the adapted ridge trace plot can be used to choose a suitable smoothing parameter for the P-spline estimator. We found that both the lag 0 storm and lag 2 low wind are significant at the 10% level of significance suggesting that the probability of asthma hospitalization on a given day is increased by occurrence of dust storms on the same day and low wind events on two days prior to the admissions.




Received from ProQuest

File Size

103 pages

File Format


Rights Holder

Priyangi Kanchana Bulathsinhala