Table 2.
Root-mean-square-error of best performing imputation method across various lengths of missing data.
| Missing data length and method | RMSEa | |
| 25 min | ||
|
|
Kalman ARIMAb | 10.5 |
|
|
Linear interpolation | 10.9 |
|
|
Kalman structural time series | 10.9 |
| 50 min | ||
|
|
Kalman ARIMA | 13.2 |
|
|
Linear interpolation | 12.3 |
|
|
Kalman structural time series | 12.3 |
| 150 min | ||
|
|
Kalman ARIMA | 14.2 |
|
|
Linear interpolation | 13.7 |
|
|
Kalman structural time series | 13.8 |
aRMSE: root-mean-square error.
bARIMA: autoregressive integrated moving average.