Details: TIMESERIES Procedure

The TIMESERIES procedure can be used to perform trend and seasonal analysis on transactional data. For trend analysis, various sample statistics are computed for each time period defined by the time ID variable and INTERVAL= option. For seasonal analysis, various sample statistics are computed for each season defined by the INTERVAL= or the SEASONALITY= option. For example, if the transactional data ranges from June 1990 to January 2000 and the data are to be accumulated on a monthly basis, then the trend statistics are computed for every month: June 1990, July 1990, ..., January 2000. The seasonal statistics are computed for each season: January, February, ..., December.

The TIMESERIES procedure can be used to form time series data from transactional data. The accumulated time series can then be analyzed using time series techniques. The data is analyzed in the following order:

1. accumulation

   ACCUMULATE= option in the ID, VAR, or CROSSVAR statement

2. missing value interpretation

   SETMISSING= option in the ID, VAR, or CROSSVAR statement

3. time series transformation

   TRANSFORM= option in the VAR or CROSSVAR statement

4. time series differencing

   DIF= and SDIF= options in the VAR or CROSSVAR statement

5. descriptive statistics

   OUTSUM= option and the PRINT=DESCSTATS option

6. seasonal decomposition

   DECOMP statement or the OUTDECOMP= option in the PROC TIMESERIES statement

7. correlation analysis

   CORR statement or the OUTCORR= option in the PROC TIMESERIES statement

8. singular spectrum analysis

   SSA statement or the OUTSSA= option in the PROC TIMESERIES statement

9. Fourier spectral analysis

   SPECTRA statement or the OUTSPECTRA= option in the PROC TIMESERIES statement

10. cross-correlation analysis

    CROSSCORR statement or the OUTCROSSCORR= option in the PROC TIMESERIES statement