Table 1.

The most successful models per stage of the different observed time series $\mathbf{r}$ are presented in the second column. Column 3 presents the percentages of the LBQ test on the different lags from $nt$ to $nt+48$ from the 27 sensors that fulfill the assumptions of independence. Column 4 presents the percentages of the LM test from the 27 sensors that fulfill the assumptions of the absence of Arch effect. The significance level used was 0.01.

Stage	Model	LBQ	LM
(a) For Method 3
Wr1A	Seasonal ARIMA$(0,1,2){(2,1,0)}_{24}$	77.00	92.52
Wr1B	Seasonal ARIMA$(0,1,0){(2,1,0)}_{24}$	3.00	44.44
CdA	Seasonal ARIMA$(0,1,0){(2,1,0)}_{24}$	25.00	77.77
CdB	Seasonal ARIMA$(0,1,2){(2,1,0)}_{24}$	18.00	18.52
Tr	Seasonal ARIMA$(0,1,3){(2,1,0)}_{24}$	11.00	3.70
HtA	Seasonal ARIMA$(1,1,3){(0,1,1)}_{24}$	22.00	22.22
HtB	Seasonal ARIMA$(0,1,3){(2,1,0)}_{24}$	0.00	37.04
Wr2	Seasonal ARIMA$(0,1,2){(2,1,0)}_{24}$	18.00	37.04
(b) For Method 4
Wr1A	A seasonal ARIMA per sensor	92.59	96.30
Wr1B	A seasonal ARIMA per sensor	51.85	59.26
CdA	A seasonal ARIMA per sensor	81.48	81.48
CdB	A seasonal ARIMA per sensor	48.15	25.93
Tr	A seasonal ARIMA per sensor	25.93	3.70
HtA	A seasonal ARIMA per sensor	59.26	29.63
HtB	A seasonal ARIMA per sensor	25.93	44.44
Wr2	A seasonal ARIMA per sensor	55.55	37.04