The log transformed larval recovery data were analyzed for effect of day, stratum, and day × stratum interaction for each grass species during two separate experimental periods.
We conducted a sensitivity analyses to determine whether the basic model results were robust to the selection of controls, comparing the aforementioned design that sampled referent times +/- 7, 14, and 21 days within 28-day strata, to a design with referent times +/- 7 and 14 days within 21-day strata.
We used 28-day strata, and compared cases with referent times 7, 14, and 21 days before or after the case, within the same stratum.
Finally, we determined the average amount of excess (flu-attributable) mortality during flu season by fitting a Poisson regression model to data on 39,420 person-day strata (12 age-sex groups × 9 years × 365 calendar days, combining the 2 extra leap days with February 28).
First, we compared the association between temperature and suicide adjusting for month and date with the unadjusted association for temperature based on 28- and 56-day strata in national capital cities (Seoul, Tokyo, and Taipei) to examine whether a 56-day stratum was appropriate.
In our sensitivity analyses, the difference in the risk estimates between the 28- and 56-day strata was smaller in the model that adjusted for month and date than the difference in those in the model that did not adjust for the two variables.
Each case was matched to 7 control days on the same day of the week during the 56-day stratum.
No significant differences were detected for stratum or day × stratum interaction effects, though stratum provided a strong indication of influencing larval recovery.
Day-of-the week effects were present in EMD-AT data (Sunday stratum, Monday to Saturday stratum) and the ED-ES data (Sunday to Monday stratum, Tuesday to Saturday stratum).
To reduce collinearity between temperature lags in our models, we modeled AT exposure as 2-day sums of lags from lags 0 7, instead of individual lags, or as four 2-day lag strata.
A priori, we split the time series into 56-day nonoverlapping strata and compared differences in exposure (temperature) between case and control days within the same stratum using a conditional logistic regression model.
