In addition to the results listed, models with lag times between independent and dependent variables were also computed.
The model space included models with lagged values of apparent temperature, sea level pressure (up to lag 20), and an air pollutant (up to lag 5).
Further, the R square statistic was greater for all age-specific models with lag compared to those without indicating better fitted estimates.
We then assessed the delayed effect through refitting the models using daily mean temperature on the day of death and the previous 13 days using distributed lag models with lag terms 0 1 days, 2 6 days, and 7 13 days.
After adjusting the admission data for yearly trends, day-of-week effects, ambient maximum temperature, and dew point temperature, we studied the associations of the pollutants in single-pollutant models with lag times of 0-4 days.
In contrast to FMOLS, augmentation of model with lags and leads of differenced regressand and regressors in DOLS helps it to overcome the problem of asymptotic endogeneity and serial correlation.
Time-dependent average exposure was modeled with lags of 0 and 5 years (continuous and in quartiles).
For example, a model with lags of 0 through 3 days from the maximum heat index yielded approximately the same deviance explained as a model with the same day and the average of 1-, 2-, and 3-day lags.
Here we see that in estimating richer models with leads and lags, there is also a large reduction in precision when geographically proximate controls are included.
