Skip to main content
. 2020 Sep 4;35(8):749–761. doi: 10.1007/s10654-020-00676-7

Table 2.

Sensitivity analysis for different values of observation and model error covariance matrices. The first two rows show number of latent infected patients and patients under treatment predicted for the 28.05.2020, the last day in the sample. The last two rows show the mean forecasting errors for treated patients and confirmed cases over the full sample for each covariance matrix configuration. The table exemplifies a bad fit with a high amount of forecasting errors when using a naive unit covariance setup

P Value 0.1 1 1 1 100 100
Q Value 0.5 100 1 10 1 10
Treatment 230,760 231,336 230,792 231,517 230,782 230,934
Infections 210,311 2,102,126 2,102,549 2,102,431 2,102,576 2,102,603
MRSFE T 888 863 901 807 829 836
MRSFE R 39,476 39,291 39,431 39,381 39,459 39,469