What exactly is an item store?

[cf8261a]

Hello,

In https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_kaplan_sect047.htm, it says you can save compiled templates in a special file called an item store. I get that. 

In https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_plm_examples02.htm, it says that the PLM procedure can be used for making post hoc inferences 'based on saved item stores'. Then it says 'The STORE statement saves the fitted model to a SAS item store named sasuser.cheese. '

So from what I'm understanding so far, an item store is some type of either file or object that allows you to store some other type of SAS object. 

Is that right? If so, I don't quite understand the usefulness of doing that? Like if I am running a regression model, what is the need to store that object in this 'item store'? 

I am new to SAS and am just having a hard time understanding what this concept is. 

[sbxkoenk · Accepted solution]

Hello,

The STORE statement requests that the (statistical) procedure save the context and results of the statistical analysis into an item store. An item store is a binary file format that cannot be modified by the user. The contents of the item store can be processed with the PLM procedure. One example of item store technology is to perform a time-consuming analysis and to store its results by using the STORE statement. At a later time you can then perform specific statistical analysis tasks based on the saved results of the previous analysis, without having to fit the model again.

I use item stores and PROC PLM regularly.

I think PLM is for Post Linear Model.

You can for example score new data using a saved (stored) model with PROC PLM.
But take care : item stores (like catalogs) are not portable between operating systems and between different encodings.

Cheers,
Koen

[cf8261a]

Okay that makes a lot of sense now. Thank you Koen!

[sbxkoenk]

I said above:

>> I think PLM is for Post Linear Model.

It is rather : 

Post-fitting (statistical analyses) Linear Models

Post-fitting can be tasks such as the following:

• testing hypotheses

• computing confidence intervals

• producing prediction plots

• scoring a new data set

Kind regards,

Koen