Chapter 11, Section 11.3.1

 

Study of Low Birth Weight Infants

Logistic Regression Model for BPD as a function of Birth Weight

 

use "bpd.dta"

 

replace weight=weight/100

(223 real changes made)

 

glm bpd weight,family(binomial 1) link(logit)

 

Iteration 0:   log likelihood = -112.47132  

Iteration 1:   log likelihood = -111.86032  

Iteration 2:   log likelihood = -111.86031  

 

Generalized linear models                          No. of obs      =       223

Optimization     : ML                              Residual df     =       221

                                                   Scale parameter =         1

Deviance         =  223.7206243                    (1/df) Deviance =  1.012311

Pearson          =  251.6649547                    (1/df) Pearson  =  1.138755

 

Variance functionV(u) = u*(1-u)                  [Bernoulli]

Link function    : g(u) = ln(u/(1-u))              [Logit]

 

                                                   AIC             =  1.021169

Log likelihood   = -111.8603121                    BIC             = -971.2643

 

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

             |                 OIM

         bpd |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

-------------+----------------------------------------------------------------

      weight |  -.4229139   .0640855    -6.60   0.000    -.5485192   -.2973087

       _cons |   4.034291   .6957851     5.80   0.000     2.670577    5.398004

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

 

 

Logistic Regression Model for BPD as a function of Birth Weight , Gestational Age, and Toxemia

 

glm bpd weight gestage toxemia,family(binomial 1) link(logit)

 

Iteration 0:   log likelihood = -103.49545  

Iteration 1:   log likelihood = -101.86315  

Iteration 2:   log likelihood = -101.85377  

Iteration 3:   log likelihood = -101.85377  

 

Generalized linear models                          No. of obs      =       223

Optimization     : ML                              Residual df     =       219

                                                   Scale parameter =         1

Deviance         =  203.7075444                    (1/df) Deviance =  .9301714

Pearson          =   289.169671                    (1/df) Pearson  =  1.320409

 

Variance functionV(u) = u*(1-u)                  [Bernoulli]

Link function    : g(u) = ln(u/(1-u))              [Logit]

 

                                                   AIC             =  .9493612

Log likelihood   = -101.8537722                    BIC             = -980.4631

 

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

             |                 OIM

         bpd |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

-------------+----------------------------------------------------------------

      weight |  -.2643577   .0812316    -3.25   0.001    -.4235687   -.1051468

     gestage |  -.3885357   .1148914    -3.38   0.001    -.6137186   -.1633527

     toxemia |  -1.343786   .6075041    -2.21   0.027    -2.534472   -.1531001

       _cons |   13.93608   2.982554     4.67   0.000     8.090383    19.78178

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