Table 1.
The comparisons of time series components between the series of mild and severe HFMD cases
| Components | Mild HFMD cases | Severe HFMD cases | ratio/differencea | p-value* |
|---|---|---|---|---|
| Long-term linear trend | ||||
| Biennial increase (%) | 124.46 (122.80, 126.14) | 91.20 (89.69, 92.74) | 1.36 (1.34,1.39) | <0.001 |
| Semi-annual cycle | ||||
| peak value | 1.04 (1.00,1.07) | 0.87 (0.82,0.93) | 1.19 (1.11,1.27) | <0.001 |
| peak time (days) | 134.7 (133.77,135.68) | 131.48 (129.83,133.13) | 3.21 (1.42,5.08) | <0.001 |
| Eight-monthly cycle | ||||
| peak value | 0.24 (0.20,0.27) | 0.21 (0.16,0.26) | 1.14 (0.88,1.52) | 0.081 |
| peak time (days) | 97.47 (91.34,103.51) | 74.78 (64.86,83.81) | 22.69 (11.03,34.24) | 0.001 |
| Annual cycle | ||||
| peak value | 1.78 (1.73,1.83) | 3.08 (2.97,3.19) | 0.58 (0.55,0.60) | <0.001 |
| peak time (days) | 183.08 (181.79,184,48) | 176.52 (175.27,177.77) | 6.56 (4.89,8.52) | <0.001 |
| Biennial cycle | ||||
| peak value | 0.40 (0.36,0.43) | 0.80 (0.75,0.86) | 0.49 (0.44,0.56) | <0.001 |
| peak time (days) | 466.49 (456.14,477.35) | 492.55 (483.99,502.11) | −26.06 (−40.96,−11.92) | <0.001 |
| First year cycle of the overall periodic curve | ||||
| Start time (days) | 36 (35,38) | 27 (25,29) | 9 (7,11) | <0.001 |
| Major peak time (days) | 155 (153,156) | 156 (155,158) | −1 (−3,0) | 0.012 |
| Major peak value | 1.23 (1.19,1.28) | 2.10 (2.00,2.19) | 0.59 (0.56,0.62) | <0.001 |
| Minor peak time (days) | 288 (284,292) | - | - | - |
| Minor peak value | 0.44 (0.42,0.47) | - | - | - |
| Second year cycle of the overall periodic curve | ||||
| Start time (days) | 391 (390,392) | 377 (375,379) | 14 (11,16) | <0.001 |
| Major peak time (days) | 510 (509,511) | 513 (512,514) | −3 (−4,−1) | <0.001 |
| Major peak value | 1.90 (1.85,1.96) | 4.20 (4.07,4.34) | 0.45 (0.43,0.47) | <0.001 |
| Minor peak time (days) | 661 (659,664) | - | - | - |
| Minor peak value | 0.42 (0.36,0.44) | - | - | - |
The 95 % confidence interval of model estimates were given in the following bracket
ato compare two series, we calculated the relative difference (i.e. the ratio of mild case to severe case) for the biennial increase and the peak value of cycles, whereas the absolute difference (i.e. mild case minus severe cases) was calculated for the start and peak timing. We applied a quasi-Poisson model to estimate the time series components in which a log function is used to link the observed values and linear predictor. Therefore, the ratio of biennial increase and peak value between two series is equivalent to the absolute difference between their related linear predictors. However, the start and peak timing will not be affected by the log link function
*The p-value was calculated to test the equality of time series components between two series, with the null hypothesis of no difference (i.e. the ratio equals 1 or difference equals to 0)