Probit Regression with SPSS

Again, unlike in SAS, SPSS models the probability of Y=1 or of Y's higher sorted value. To fit a probit regression, use: 

  probit r of n with x1 x2
   /model probit.
Example 17: SPSS Probit Regression in PROBIT procedure

Using the data in Example 1, you can use:

  compute n = 1.
  execute.
  probit r of n with t 
   /model probit
   /print none.

The resulting SPSS output will be:

* * * * * * * * * * * *  P R O B I T    A N A L Y S I S  * * * * * * * * * * * *

 Parameter estimates converged after 13 iterations.
 Optimal solution found.

 Parameter Estimates (PROBIT model:  (PROBIT(p)) = Intercept + BX):

           Regression Coeff.  Standard Error     Coeff./S.E.

   T                 -.03908          .01142        -3.42024


                   Intercept  Standard Error  Intercept/S.E.

                     2.80035          .33162         8.44443

  Pearson  Goodness-of-Fit  Chi Square =    352.383    DF = 385   P =  .882

  Since Goodness-of-Fit Chi square is NOT significant, no heterogeneity
  factor is used in the calculation of confidence limits.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

SPSS PROBIT procedure supports the inclusion of categorical variables as explanatory variables. However, it only accepts numerically coded categorical variables. If your categorical variable is string, you need to reassign string values to numerical values in a new variable before running PROBIT.

Example 18: SPSS Probit Regression in PROBIT procedure (categorical regressors)

Using the data in Example 7, you can use:

  autorecode variables = drug / into drug2.
  probit r of n by drug2(1 5) with x
   /model probit
   /print none
   /criteria iterate(20) steplimit(.1).

where AUTORECORD reassigns the string values A, B, C, D, and E of the variable DRUG to the consecutive integers 1,2,3,4, and 5 in the new variable DRUG2.

The resulting SPSS output will be:

  
* * * * * * * * * * * *  P R O B I T    A N A L Y S I S  * * * * * * * * * * * *

DATA  Information

         18 unweighted cases accepted.
          0 cases rejected because of out-of-range group values.
          0 cases rejected because of missing data.
          0 cases are in the control group.

Group Information

    DRUG2     Level  N of Cases    Label
                  1           3        A
                  2           4        B
                  3           4        C
                  4           3        D
                  5           4        E


MODEL Information

      ONLY Normal Sigmoid is requested.

 Parameter estimates converged after 18 iterations.
 Optimal solution found.

 Parameter Estimates (PROBIT model:  (PROBIT(p)) = Intercept + BX):

           Regression Coeff.  Standard Error     Coeff./S.E.

   X                 1.15886          .43833         2.64378


                   Intercept  Standard Error  Intercept/S.E.  DRUG2

                    -1.51849          .32692        -4.64487  A
                    -1.03837          .26286        -3.95027  B
                    -2.04067          .37455        -5.44829  C
                     -.31766          .33551         -.94678  D
                      .19031          .24926          .76350  E

  Pearson  Goodness-of-Fit  Chi Square =      4.383    DF = 12   P =  .975

  Since Goodness-of-Fit Chi square is NOT significant, no heterogeneity
  factor is used in the calculation of confidence limits.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Next: Models for Multiple Outcomes
Prev: Probit Regression with SAS
Up: Probit Regression