Logistic regression

• As an example, we take data from the BHPS, wave 5:
  • Dependent: education; A-level-plus versus less than A-level
  • Sex
  • Age (grouped into 10-year bands, highest 70+)
  • Vote (as before).

• We can fit this as a conventional logistic regression:

  logistic regression hied with esex ageg vote 
    /cat=vote ageg 
    /con(vote)=ind/con(ageg)=ind.
  

  HIED is the dependent variable, VOTE and AGEG are declared categorical, and simple dummy variable (indicator) contrasts are requested.

  Variable           B      S.E.     Wald    df      Sig       R   Exp(B)
  
  ESEX          -.3597     .0520  47.8442     1    .0000  -.0660    .6979
  AGEG                           1203.612     6    .0000   .3367
   AGEG(1)      2.4024     .1463 269.6315     1    .0000   .1596  11.0492
   AGEG(2)      2.9158     .1037 790.2274     1    .0000   .2739  18.4627
   AGEG(3)      2.2740     .0913 619.7681     1    .0000   .2425   9.7181
   AGEG(4)      1.7560     .0902 378.9330     1    .0000   .1894   5.7890
   AGEG(5)      1.1135     .0908 150.4871     1    .0000   .1189   3.0450
   AGEG(6)       .4487     .0985  20.7411     1    .0000   .0422   1.5663
  VOTE                           212.9447     3    .0000   .1403
   VOTE(1)      1.0285     .0889 133.9467     1    .0000   .1120   2.7968
   VOTE(2)       .2291     .0784   8.5261     1    .0035   .0249   1.2574
   VOTE(3)       .8494     .1040  66.6860     1    .0000   .0785   2.3382
  Constant      -.9208     .1289  51.0232     1    .0000