IBM Support

Reference Category in Multinomial Logistic Regression

Troubleshooting


Problem

I am running the Multinomial Regression procedure (NOMREG) command in SPSS/PASW Statistics. My dependent variable has four levels. I understand that the last category (the highest numeric value or the last alphabetic value) is the reference category for the dependent variable by default. How can I specify one of the first three categories as the reference category?

Resolving The Problem

By default, Multinomial Logistic Regression (NOMREG) uses the last (highest) category level as the reference category for the dependent variable (DV). However, you can choose an alternate reference category for the DV. In the main Multinomial Logistic Regression dialog, paste the dependent variable into the "Dependent Variable" box. You can then click the "Reference Category" below that box. In the Reference Category dialog, you can click the button for "First category", "Last Category" or "Custom". Choose "Custom" if you want an intermediate category as reference category, followed by entering the value for the reference category in the "Value" box below the "Custom" button.
Also note that the order of the levels can be specified as ascending, descending, or as the order in which the category levels are encountered in the data. The Ascending and Descending choices are available in the Reference Category dialog, but the Data order must be specified in a syntax command.

Here is an example of a NOMREG syntax command which specifies that the first DV category should be the reference category. DEPVAR is the DV, A and B are factors, and X is a covariate.

NOMREG
depvar (BASE=FIRST) BY a b WITH x
/CRITERIA CIN(95) DELTA(0) MXITER(100) MXSTEP(5) CHKSEP(20) LCONVERGE(0)
PCONVERGE(0.000001) SINGULAR(0.00000001)
/MODEL
/INTERCEPT INCLUDE
/PRINT PARAMETER SUMMARY LRT .

To specify an arbitrary level, specify the value after BASE, for example (BASE=2). If the dependent variable is a string, give the value in quotation marks, for example (BASE='no'). To explicitly specify the last category as the reference use (BASE=LAST).

Similarly, the order in which the category levels are displayed in the table of Parameter Estimates may be controlled with ORDER. The choices for ORDER are ASCENDING, DESCENDING, or DATA, with ASCENDING the default. BASE uses the order specified in determining what the first or last level is.

For example:

NOMREG depvar (BASE=LAST ORDER=ASCENDING) BY a b WITH x
/MODEL
/PRINT=PARAMETER SUMMARY LRT.

This gives exactly the same results as if nothing were specified for BASE or ORDER, since we have told SPSS to make the same choices it would have if we had done nothing. In contrast:

NOMREG depvar (BASE=FIRST ORDER=DATA) BY a b WITH x
/MODEL
/PRINT=PARAMETER SUMMARY LRT.

This will use the first level encountered in the data as the reference, and also display the parameter estimates in the same order as the levels are found in the data.

Run the modified syntax by clicking the black triangle button on the toolbar, by right-clicking in the Syntax window and choosing Run current..., or by using one of the choices in the Run... menu.

At this time, it is not possible to specify the reference level of the independent factors. This must still be accomplished through recoding, so that the desired category is last.

[{"Product":{"code":"SSLVMB","label":"IBM SPSS Statistics"},"Business Unit":{"code":"BU059","label":"IBM Software w\/o TPS"},"Component":"Not Applicable","Platform":[{"code":"PF033","label":"Windows"}],"Version":"Not Applicable","Edition":"","Line of Business":{"code":"LOB10","label":"Data and AI"}}]

Historical Number

26615

Document Information

Modified date:
16 April 2020

UID

swg21479637