IBM Support

Multiple Imputation Warning that model contains more than 100 parameters



I used the Analyze>Multiple Imputation>Analyze Patterns... menu before completing the data with Analyze>Multiple Imputation>Impute Missing Data Values... However, all I get is a table of titled Warnings: The imputation model for [variable] contains more than 100 parameters. No missing values will be imputed. Reducing the number of effects in the imputation model, by merging sparse categories of categorical variables, changing the measurement level of ordinal variables to scale, removing two-way interactions, or specifying constraints on the roles of some variables, may resolve the problem. Alternatively increase the maximum number of parameters allowed on the MAXMODELPARAM keyword of the IMPUTE subcommand. This command is not executed. How can I impute missing values for my data?

Resolving The Problem

Check the type of Measure for each variable: this can be done conveniently by going to the Data Editor, and choosing the Variable View tab in the lower left. The Analyze Patterns dialog performs a scan of the data (as do several other procedures). As a result of this scan, the Measure of some variables may have been set to Nominal when there are up to 20 distinct values appearing in that variable. For example, variables X and Y might be scale variables which happen to have more than 10 values each, but fewer than 20. Multiple Imputation will use Multinomial Logistic Regression as the model for Y with X (and any scale variables present ) as predictors. There will be 10 or more parameters for X (plus any for the scale variables), multiplied by the number of values of Y minus 1. This product can easily be in excess of 100. (Similarly, a Multinomial Logistic regression model for X will be fit using Y as one of the predictors. It too will likely have more than 100 parameters.)

If the variables were not intended to be Nominal, but it just happened that 10 to 20 values occurred in the data, simply change their Measure to Scale in the Data Editor, Variable View tab.

If some of the variables are Nominal, but there are many categories, it is a good idea to use Analyze>Descriptive Statistics>Frequencies... to check the distribution of values. If there are categories which don't occur very often, it is probably best to combine them with other similar values if possible.

Specifying constraints on the role of variables, or (using SPSS command syntax) customizing the model for each variable to be imputed could also resolve the problem.

But be careful about simply increasing MAXMODELPARAM in this circumstance, unless you are sure that the models are appropriate. It is likely that Multiple Imputation will take a very long time to finish. If you do want to adjust this and are not familiar with command syntax, once you've specified the desired analysis in the Impute Missing Data Values dialog box, click Paste. A MULTIPLE IMPUTATION command will be pasted into a syntax window. Look for the IMPUTE subcommand. Place the cursor at the end of that line and type MAXMODELPARAM= and give the desired number.

[{"Product":{"code":"SSLVMB","label":"SPSS Statistics"},"Business Unit":{"code":"BU053","label":"Cloud & Data Platform"},"Component":"Not Applicable","Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"Not Applicable","Edition":"","Line of Business":{"code":"LOB10","label":"Data and AI"}}]

Historical Number


Document Information

Modified date:
16 April 2020