The SURVEYIMPUTE Procedure

OUTPUT Statement

  • OUTPUT <OUT=SAS-data-set> <OUTJKCOEFS=SAS-data-set> <keyword=name …keyword=name> ;

The OUTPUT statement creates a SAS data set that contains the imputed data. You must use the OUTPUT statement to store the imputed data in a SAS data set. If you use multiple OUTPUT statements, then PROC SURVEYIMPUTE uses only the first OUTPUT statement and ignores the rest. The OUTPUT OUT= data set contains all the variables from the DATA= input data set, imputed values for missing values for the variables in the VAR statement, and some observation-level quantities. These quantities can include the fractionally adjusted weights, replicate weights, recipient numbers, and donor identifications.

You can specify the following in the OUTPUT statement:

OUT=SAS-data-set

names the output data set. If you use the OUTPUT statement but omit the OUT= option, then the output data set is named by using the DATAn convention. For more information, see the section OUT= Output Data Set.

OUTJKCOEFS=SAS-data-set

names a SAS data set that contains the jackknife coefficients for VARMETHOD=JACKKNIFE .

keyword <=name>

specifies the quantities to include in the output data set and optionally names the new variables that contain the quantities. Specify a keyword for each desired quantity (see the following list of keywords), optionally followed by an equal sign and a variable name to contain the quantity. If you specify a keyword without a variable name, then PROC SURVEYIMPUTE uses default names. You can specify the following keywords:

DONORID<=name>

requests a name for the identification variable for the donor units. If you do not specify this keyword, the donor IDs are not saved in the output data set. If you specify this keyword but do not specify a name, then the donor IDs are stored in a new variable named DonorID. This keyword is available only when METHOD=HOTDECK .

FRACTIONALWEIGHTS=name

includes the fractional weights of donor cells in the output data set and specifies the corresponding variable name. If you do not specify this keyword, the fractional weights are not saved in the output data set. This keyword is available only when METHOD=FEFI .

IMPADJWEIGHTS=name

includes the imputation-adjusted weights in the output data set and specifies the corresponding variable name. The imputation-adjusted weights are computed by multiplying the base weights by the fractional weights. If you do not specify this keyword, then the imputation-adjusted weights are stored in a new variable named ImpWt. This keyword is available only when METHOD=FEFI .

IMPSTATUS=name

includes an imputation status index with the values shown in Table 110.3.

Table 110.3: Imputation Status Index

Index

Imputation Status

0

All items are observed

1

All missing items are imputed

2

No missing items are imputed

3

Some missing items are imputed but some missing items are not imputed

4

Invalid observation; observation is not used in the imputation


OBSID=name

includes an index variable to contain the unique numeric identification of every unit from the input data set in the output data and specifies the corresponding variable name. If you do not specify this keyword, then the default unit ID is stored in a new variable named UnitID. The OBSID= option is not applicable when the ID statement is specified.

RECIPIENT=name

includes the recipient index in the output data set and specifies the corresponding variable name. The recipient index can be 0 (which indicates a nonmissing unit) or any positive integer (which represents multiple donor units for a recipient unit). If you do not specify this keyword, then the recipient index is stored in a new variable named Recipient.