topic Re: time series in SAS Data Science

time series

Lopa2016 — Tue, 09 Jan 2018 17:11:03 GMT

Hi ,

I have monthly data sets which look like this:

dismonth	pross	Actual
JUL15	AUG15	4415822.77
JUL15	SEP16	33578.75
JUL15	NOV15	408408.93
JUL15	JUL15	3347038.23
JUL15	JUN16	59340
JUL15	OCT17	5526.5
JUL15	MAR16	112341.58
JUL15	JUL17	4363
JUL15	SEP15	1611128.75
JUL15	APR17	4426
JUL15	JAN16	140840.5
JUL15	JAN17	11595.75
JUL15	OCT16	21883.75
JUL15	JUL16	61080.5
JUL15	OCT15	696127.06
JUL15	DEC17	2350
JUL15	APR16	89935.1
JUL15	SEP17	2544.5
JUL15	JUN17	4731.5
JUL15	MAR17	15919
JUL15	DEC16	13774
JUL15	DEC15	300524.3
JUL15	AUG16	58034.25
JUL15	MAY16	80908.5
JUL15	NOV17	9548.4
JUL15	AUG17	7095.36
JUL15	FEB16	157373.5
JUL15	MAY17	3204
JUL15	FEB17	8664.25
JUL15	NOV16	15808

I want to forecast for the next 5 years but when I use proc arima few of the forecasts are negative & look incorrect & there are a few warnings :

Warning: There are gaps in the interval for observation 2 according to ID variable PROSS_DATE.

Can some one help out?

Thanks in advance.

Re: time series

Lopa2016 — Tue, 09 Jan 2018 19:31:00 GMT

I believe proc expand is the solution

Re: time series

DougWielenga — Fri, 19 Jan 2018 17:44:28 GMT

It is not clear to me what you are trying to do. You said you have monthly data sets that look like your example (presumably DISMONTH which in your data only shows observations for July 2015), so what is the PROSS date? Are you trying to analyze the monthly files separately or was the data you showed just one month from a large set of monthly data which has many observations? Either way, the EXPAND procedure can help you interpolate any missing values and or accumulate your data to the desired level. It would be easier if you sorted the data by date as needed to be able to more easily see patterns. There seems to be some wild fluctuations in ACTUAL but it is hard to see if any patterns are present when the data is not sorted. Were you able to get what you needed using the EXPAND procedure?

Cordially,

Doug