No Subject indicator in PROC SSM for State-Space Model

[ccwangxi]

I would like to apply the state-space model to my longitudinal dataset. The datasets consist of daily-collected data in 14 weeks from about 200 subjects, similar to the one provided in the SAS example https://documentation.sas.com/doc/en/etscdc/14.3/etsug/etsug_ssm_examples04.htm. However, the subject ID is not included in the code. So, does PROC SSM incorporate the subject ID information, or is it similar to the classic time series analysis (for one subject -- ignoring the subject ID)?

Would anyone know the reference code for an implementation? Thank you!

As a reference:

 proc ssm data=dataset;
    id time;
    trend growth(ps(2));
    irregular wn;/*wn is white noise, residual error*/
    model y = treatment growth wn;
    eval y_predict = treatment + growth;
    output out=For;
 quit;

[sbxkoenk]

Hello,

Do you want a different SSM analysis for every subject?

In that case

, sort your dataset by subject and time

, and then add a by statement to your code.

by subject_id;
id time;

Good luck,

Koen

[ccwangxi]

Hi Koen,

Thank you for your reply! I knew that we can get different SSM analyses for every subject using "by subject_id;". But is there a possible way to get the population-level fixed effects coefficients? 

Best,

Xi

[sbxkoenk · Accepted solution]

Hello,

if you have time series for several subjects, you have time-series cross-sectional data, aka panel data.

For panel data you can use PROC TSCSREG, PROC PANEL, PROC CPANEL, PROC MIXED(, PROC GLIIMIX) or PROC SSM indeed.

See here :

SAS 9.4 / Viya 3.5 -- SAS/ETS User's Guide
The SSM Procedure
Example 34.2 Panel Data: Random-Effects and Autoregressive Models

https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/etsug/etsug_ssm_examples02.htm?homeOnFail

Koen