AIC & SBC negative

[azhlla]

Hi so I do research to analyze the impact of earthquake and COVID-19 towards OCC stars hotel with ARIMA intervention and got AR(1) for the ARIMA and the results have AIC n SBC negative, so do the first intervention and second intervention. the intervention orders for first intervention are b=0 s=0 and r=0, how do I solve this? Thank you in advance🥺🙏

[Rick_SAS · Accepted solution]

>  how do I solve this?

It is not clear to me what you are trying to solve. AIC and SBC can both be negative. That (by itself) does not indicate that something is wrong with the model. For example, in PROC GLMSELECT, the definitions for AIC and SBC include a term of the form log(SSE / n). If SSE < n, then this quantity is negative.

Please post the SAS procedure statements that you are using to specify the model. That will give us more information.

[azhlla]

Thank you so much for the explanation. My box cox result for the lambda is 1.5 so transform the data. Here is my procedure:

/*IMPORT DATA*/
LIBNAME PRA XLSX "/home/u61344338/TPK_mul/PRA.xlsx";
/*BOX-COX TEST*/
DATA PRA.HB; SET PRA.HB; TRHBPI=(HBPI)**1.5; Y=0; RUN;
PROC TRANSREG MAXITER=0 NOZEROCONSTANT;
MODEL BOXCOX(TRHBPI)=IDENTITY(Y);
RUN;

/*ADF TEST*/
PROC ARIMA DATA=PRA.HB;
IDENTIFY VAR=TRHBPI NLAG=24 STATIONARITY=(ADF);
RUN;

/*ARIMA PRAINTERVENTION MODEL*/
/*ARIMA100 TRHBPI*/
PROC ARIMA DATA=PRA.HB PLOTS
(ONLY)=(SERIES(CORR CROSSCORR) RESIDUAL(CORR NORMAL)
FORECAST(FORECASTONLY));
IDENTIFY VAR=TRHBPI;
ESTIMATE P=1 METHOD=CLS;
FORECAST LEAD=19 BACK=0 ALPHA=0.05 OUT=PRA.HBF100 PRINTALL;
OUTLIER;
RUN;
PROC UNIVARIATE DATA=PRA.HBF100 NORMAL;
VAR RESIDUAL;
RUN;

/* FIRST ARIMA INTERVENTION*/

LIBNAME AI1 XLSX "/home/u61344338/TPK_mul/AI1.xlsx";
/*AI1 b=0 s=0*/
PROC ARIMA DATA=AI1.HB;
IDENTIFY VAR=TRHBI1 CROSSCORR=(s) NLAG=18 NOPRINT;
ESTIMATE P=(1) INPUT=(0$ (0) s);
RUN;

FORECAST LEAD=22 OUT=AI1.HBF PRINTALL;
RUN;

PROC UNIVARIATE DATA=AI1.HBF NORMAL;
VAR RESIDUAL;
RUN;

/*SECOND ARIMA INTERVENTION*/
/*AI2 b=0 s=7*/

LIBNAME AI2 XLSX "/home/u61344338/TPK_mul/AI2.xlsx";

PROC ARIMA DATA=AI2.HB;
IDENTIFY VAR=TRHBI2 CROSSCORR=(s1 s2) NLAG=24 NOPRINT;
ESTIMATE P=(1) INPUT=(0$ (7) s2 0$ (0) s1);
RUN;

FORECAST LEAD=12 OUT=AI2.HBF PRINTALL;
RUN;

PROC UNIVARIATE DATA=AI2.HBF NORMAL;
VAR RESIDUAL;
RUN;

Is there anything wrong with it?