topic Re: time series: how to account for 5 days a week data (instead of 7) in Statistical Procedures

time series: how to account for 5 days a week data (instead of 7)

Elisabeth1 — Mon, 20 Jan 2020 19:40:39 GMT

Hi,

I am using the SAS studio forecasting procedures which generate the following code to test for White Noise.

I am using a series spanning from 2000-2019 with daily data; HOWEVER the data that I am using is reported only 5 days a week (Monday-Friday); the code generated is below and shows "seasonality=7* (i.e. 7 days of the week).

I have tried to just adjust it to "seasonality=5" however, this does not seem to yield the right results.

How should the code be adjusted to account for only 5 days a week.

ods noproctitle;
ods graphics / imagemap=on;

proc sort data=PG2.DATASET_2000_2019 out=Work.preProcessedData;
by date;
run;

proc timeseries data=Work.preProcessedData seasonality=7 plots=(series corr acf
wn);
id date interval=day;
var Spread / accumulate=none transform=none dif=0 sdif=0;
ods exclude ACFNORMPlot;
ods exclude WhiteNoiseProbabilityPlot;
run;

proc delete data=Work.preProcessedData;
run;

Re: time series: how to account for 5 days a week data (instead of 7)

novinosrin — Mon, 20 Jan 2020 19:50:28 GMT

Hi @Elisabeth1 Can the interval be shifted by 2 days?

For example: id date interval=week.2;

Disclaimer: I am only taking a guess

Re: time series: how to account for 5 days a week data (instead of 7)

mkeintz — Mon, 20 Jan 2020 19:51:14 GMT

You have a statement:

ID DATE INTERVAL=DAY;

Is there a "weekday" interval in proc timeseries? Then you might try:

ID DATE INTERVAL=WEEKDAY;

Re: time series: how to account for 5 days a week data (instead of 7)

PGStats — Mon, 20 Jan 2020 20:22:15 GMT

According to the documentation of SAS time series procedures, what @mkeintz suggested should work:

https://documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_tsdata_sect030.htm&docsetVersion=14.3&locale=en