| Thanks for that, Timothy.  I have two further questions.
|
| 2. One of the chief complaints about SPSS is that the output
| for many procedures is unwieldy.  It often contains a lot of
| stuff that (a lot of people think) ought to be optional
| rather than automatic (e.g., multivariate results for
| repeated measures ANOVA).  And on the flip side, there are
| things that perhaps ought to be automatic that are optional
| (e.g., cell means for ANOVA models obtained via GLM).  Can
| anyone comment on how Stata's output compares to that of SPSS?
|
Stata has a different approach than SPSS to output. here is a command for regression and all of the output:
 regress  mpg foreign displacement weight, beta
 
      Source |       SS       df       MS              Number of obs =      74
-------------+------------------------------           F(  3,    70) =   34.70
       Model |  1600.12629     3  533.375429           Prob > F      =  0.0000
    Residual |  1076.03588    70  15.3719411           R-squared     =  0.5979
-------------+------------------------------           Adj R-squared =  0.5807
       Total |  2676.16216    73  36.6597556           Root MSE      =  3.9207
 
------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|                     Beta
-------------+----------------------------------------------------------------
     foreign |  -1.972259   1.272818    -1.55   0.126                -.1499012
displacement |  -.0042919   .0115094    -0.37   0.710                -.0650984
      weight |  -.0061926   .0013333    -4.64   0.000                -.7948907
       _cons |   41.56386   2.686682    15.47   0.000                        .
------------------------------------------------------------------------------
At this point there are any number of specialized things you might want to do. Soooo, Stata gives you basic output in a condensed way and leaves it to you if you want more. Here
is an example using logistic regression with additional features. You can get the additional features by entering the command "findit listcoef
 
. logit foreign  mpg weight length displacement
 
Iteration 0:   log likelihood =  -45.03321
Iteration 1:   log likelihood = -27.876913
Iteration 2:   log likelihood = -23.541419
Iteration 3:   log likelihood = -21.694712
Iteration 4:   log likelihood = -20.779084
Iteration 5:   log likelihood = -20.504477
Iteration 6:   log likelihood = -20.478582
Iteration 7:   log likelihood = -20.478309
Iteration 8:   log likelihood = -20.478308
 
Logit estimates                                   Number of obs   =         74
                                                  LR chi2(4)      =      49.11
                                                  Prob > chi2     =     0.0000
Log likelihood = -20.478308                       Pseudo R2       =     0.5453
 
------------------------------------------------------------------------------
     foreign |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         mpg |  -.1728742   .0939972    -1.84   0.066    -.3571053    .0113568
      weight |   .0012032   .0029597     0.41   0.684    -.0045977     .007004
      length |   .0071799   .0706211     0.10   0.919    -.1312349    .1455946
displacement |  -.0796186   .0363671    -2.19   0.029    -.1508968   -.0083404
       _cons |   10.05333   9.018975     1.11   0.265    -7.623534     27.7302
------------------------------------------------------------------------------
 
note: 1 failure and 0 successes completely determined.
 
. listcoef
 
logit (N=74): Factor Change in Odds
 
  Odds of: Foreign vs Domestic
 
----------------------------------------------------------------------
     foreign |      b         z     P>|z|    e^b    e^bStdX      SDofX
-------------+--------------------------------------------------------
         mpg |  -0.17287   -1.839   0.066   0.8412   0.3511     6.0547
      weight |   0.00120    0.407   0.684   1.0012   2.5474   777.1936
      length |   0.00718    0.102   0.919   1.0072   1.1734    22.2663
displacement |  -0.07962   -2.189   0.029   0.9235   0.0007    91.8372
----------------------------------------------------------------------
 
. listcoef, percent
 
logit (N=74): Percentage Change in Odds
 
  Odds of: Foreign vs Domestic
 
----------------------------------------------------------------------
     foreign |      b         z     P>|z|      %      %StdX      SDofX
-------------+--------------------------------------------------------
         mpg |  -0.17287   -1.839   0.066    -15.9    -64.9     6.0547
      weight |   0.00120    0.407   0.684      0.1    154.7   777.1936
      length |   0.00718    0.102   0.919      0.7     17.3    22.2663
displacement |  -0.07962   -2.189   0.029     -7.7    -99.9    91.8372
----------------------------------------------------------------------
 
.
The listcoef gives you the exponentiated B's and what they are for a one standard deviation change in the predictor. A one standard deviation shift in weight more than doubles the odds it is a domestic car. The listcoef, percent shows you that their is a 155% increase in the odds it is a domestic car for a one standard deviation shift in weight.  The listcoef is just one example of an extension on basic logistic regression to help you interpret it.
 
Alan Acock
 

Reply via email to