concentrations for only the first two years of data, for which
the daily mortality counts appeared to be more reliable. Because
of the amount of missing data for measured PM10, it
was not feasible to estimate the daily mortality model for
only 1992 and 1993 using the measured PM10 concentrations.
We also examined the models after including binary
variables representing the day of the week, and after using
smoothers for meteorologic covariates. All of the models were
checked for autocorrelation and overdipsersion, and when
appropriate, corrections were applied.