Constraints (Multiple Imputation)

The Constraints tab allows you to restrict the role of a variable during imputation and restrict the range of imputed values of a scale variable so that they are plausible. In addition, you can restrict the analysis to variables with less than a maximum percentage of missing values.

Scan of Data for Variable Summary. Clicking Scan Data causes the list to show analysis variables and the observed percent missing, minimum, and maximum for each. The summaries can be based on all cases or limited to a scan of the first n cases, as specified in the Cases text box. Clicking Rescan Data updates the distribution summaries.

Define Constraints

  • Role. This allows you to customize the set of variables to be imputed and/or treated as predictors. Typically, each analysis variable is considered as both a dependent and predictor in the imputation model. The Role can be used to turn off imputation for variables that you want to Use as predictor only or to exclude variables from being used as predictors (Impute only) and thereby make the prediction model more compact. This is the only constraint that may be specified for categorical variables, or for variables that are used as predictors only.
  • Min and Max. These columns allow you to specify minimum and maximum allowable imputed values for scale variables. If an imputed value falls outside this range, the procedure draws another value until it finds one within the range or the maximum number of draws is reached (see Maximum draws below). These columns are only available if Linear Regression is selected as the scale variable model type on the Method tab.
  • Rounding. Some variables may be used as scale, but have values that are naturally further restricted; for instance, the number of people in a household must be integer, and the amount spent during a visit to the grocery store cannot have fractional cents. This column allows you to specify the smallest denomination to accept. For example, to obtain integer values you would specify 1 as the rounding denomination; to obtain values rounded to the nearest cent, you would specify 0.01. In general, values are rounded to the nearest integer multiple of the rounding denomination. The following table shows how different rounding values act upon an imputed value of 6.64823 (before rounding).
Table 1. Rounding results
Rounding Denomination Value to which 6.64832 is rounded
10 10
1 7
0.25 6.75
0.1 6.6
0.01 6.65

Exclude variables with large amounts of missing data. Typically, analysis variables are imputed and used as predictors without regard to how many missing values they have, provided they have sufficient data to estimate an imputation model. You can choose to exclude variables that have a high percentage of missing values. For example, if you specify 50 as the Maximum percentage missing, analysis variables that have more than 50% missing values are not imputed, nor are they used as predictors in imputation models.

Maximum draws. If minimum or maximum values are specified for imputed values of scale variables (see Min and Max above), the procedure attempts to draw values for a case until it finds a set of values that are within the specified ranges. If a set of values is not obtained within the specified number of draws per case, the procedure draws another set of model parameters and repeats the case-drawing process. An error occurs if a set of values within the ranges is not obtained within the specified number of case and parameter draws.

Note that increasing these values can increase the processing time. If the procedure is taking a long time, or is unable to find suitable draws, check the minimum and maximum values specified to ensure they are appropriate.

How To Specify Constraints for Multiple Imputation

This feature requires the Missing Values option.

  1. From the menus choose:

    Analyze > Multiple Imputation > Impute Missing Data Values...

  2. In the Impute Missing Data Values dialog box, click the Constraints tab.