Timeseries without Proc Timeseries or Proc Arima

[GS2]

Hello,

I am using SAS 9.4 but my license does not support proc timeseries or proc arima, are there any good examples of how to perform a timeseries analysis without using these two procedures? For background, I have a dataset looking at procedures from 2010-2022 and I am looking at the rate of rate of an event pre/post implementation of an intervention. Thank you for any help! 

[Kurt_Bremser]

Please post an example of your data, in text form, preferably as DATA step code with datalines.

Then show what result you expect from this.

[sbxkoenk]

PROC GEE, PROC GENMOD, PROC GLIMMIX can be tried.

• Usage Note 70498: Interrupted time series analysis: Assess the effects of an intervention on subjects observed longitudinally
  https://support.sas.com/kb/70/498.html
• Home > Analytics > Stat Procedures >
  Interrupted time series using prox glimmix or proc genmod with a binary response variable
  https://communities.sas.com/t5/Statistical-Procedures/Interrupted-time-series-using-prox-glimmix-or-...

[EDIT two weeks later]

You might also use PROC MIXED for this. This paper has some information and examples:
SAS Global Forum 2020 -- Paper 4674-2020
Time After Time: Difference-in-Differences and Interrupted Time Series Models in SAS®
E Margaret Warton, Kaiser Permanente Northern California Division of Research

https://www.sas.com/content/dam/SAS/support/en/sas-global-forum-proceedings/2020/4674-2020.pdf

Koen

[mkeintz]

I admit ignorance of PROC GEE, GENMOD, and GLIMMIX.

But some timeseries analyses will depend crucially on accommodating serial autocorrelation of error terms.  Mostly in econometrics models.

Is that a relevant concern for this study?  If so, do these proc's provide a way of finding or controlling for this type of autocorrelation?

[sbxkoenk]

---

---

---

@mkeintz wrote:

I admit ignorance of PROC GEE, GENMOD, and GLIMMIX.

But some timeseries analyses will depend crucially on accommodating serial autocorrelation of error terms.  Mostly in econometrics models.

Is that a relevant concern for this study?  If so, do these proc's provide a way of finding or controlling for this type of autocorrelation?

---

Yes indeed, because in those procedures you can do "Repeated Measures Data Analysis". So there is no longer the assumption of independent data points (like in PROC REG and PROC GLM).
In PROC GLIMMIX, for example, the R matrix is the variance-covariance matrix of the residuals and typically it is of type AR(1) or TOEP (Toeplitz). Note: The random effects in the study and the error term are assumed to be independent random variables.
So to some extent you can correctly handle longitudinal data (for many subjects), but for real econometric analyses these procedures obviously fall short.

Koen