The RESTORE statement specifies an item store that provides spatial correlation model input for the PROC KRIGE2D prediction tasks. An item store is a binary file defined by the SAS System. You cannot modify the contents of an item store. The KRIGE2D procedure can use only item stores created by PROC VARIOGRAM.

Item stores enable you to use saved correlation models without having to repeat specification of these models in the MODEL statement. In principle, an item store contains the chosen model from a model fitting process in PROC VARIOGRAM. If more than one model form is fitted, then all successful fits are included in the item store. In this case, you can choose any of the available models to use for prediction with the STORESELECT(MODEL=) option in the MODEL statement. Successfully fitted models might include questionable fits, which are so flagged when you specify the INFO option to display model names.

The store-name is a usual one- or two-level SAS name, as for SAS data sets. If you specify a one-level name, then the item store resides in the WORK library and is deleted at the end of the SAS session. Since item stores are often used for postprocessing tasks, typical usage specifies a two-level name of the form libname.membername.

When you specify the RESTORE statement, the default output contains some general information about the input item store. This information includes the store name, label (if assigned), the data set that was used to create the store, BY group information, the procedure that created the store, and the creation date.

You can specify the following option in the RESTORE statement after a slash (/):

When you specify an input item store with the RESTORE statement in PROC KRIGE2D, all the DATA= input data set variables must match input item store variables. If there are BY groups in the input DATA= set or in the input RESTORE variables, then PROC KRIGE2D handles the different cases as follows:

• If both PROC KRIGE2D has BY groups and the RESTORE statement has BY groups, then the analysis variables must match. This matching assumes implicitly that in each BY group of PROC KRIGE2D and the item store, the corresponding set of observations and correlation model comes from the same random field. This assumption is valid if you use the same data set, first in PROC VARIOGRAM to fit a model and save it in the item store, and then in PROC KRIGE2D to perform predictions with the resulting correlation models.

• If PROC KRIGE2D has BY groups but the item store does not, then the item store is accepted only if the procedure and the item store analysis variables match. In this case, the same item store model choice iterates across the BY groups of the input data. You are advised to proceed with caution: each BY group in the input DATA= set corresponds to a different realization of a random field. Hence, by using the same correlation model for prediction purposes, you implicitly assume that all these different realizations are instances of the same random field.

• If PROC KRIGE2D has no BY groups but the item store does, then the item store is rejected.