Example L6 Count response model

In this example we use data (drvisits.dat) from the German Socio-Economic Panel (SEOP) on the self reported number of visits by women to the doctor, just before a major health care reform in 1997 and just after the reform. The health reform raised prescription charges by 200% and imposed limits on fees charged by doctors. This data was analysed by Winkelmann (2004), Rabe-Hesketh and Skrondal (2005, Ch 6) who note that it is interesting to establish if the health reforms tended to reduce visits by the women to the doctor. Following Rabe-Hesketh and Skrondal (2005, Ch 5) we consider the subset of women who were working full time in the 1996 and 1998 panels of SEOP.

Rabe-Hesketh and Skrondal (2005, Ch 5) show that fewer than 50% of the women provide data for both the 1996 and 1998 waves. For the purposes of this analysis and like Rabe-Hesketh and Skrondal (2005, Ch 5) we treat the missing data as ignorable and use all the observed responses.

References

Rabe-Hesketh, S., and Skrondal, A., (2005), Multilevel and Longitudinal Modelling using Stata, Stata Press, Stata Corp, College Station, Texas.

Winkelmann, R., (2004), Healthcare reform and thye number of doctor visits: an econometric analysis, Journal of Applied Econometrics, 19, 455-472.

Data description

Number of observations (rows): 2227

Number of variables (columns): 10

The subset of variables we use are:

id=person identifier;

numvisits=self reported number of visits to the doctor during the previous 3 months;

age=age in years (20 to 60);

educ=education in years;

married=1 if the woman was married, 0 otherwise;

badh=1 if the self reported state of health was very poor or poor, 0 otherwise;

loginc=logarithm of household income in 1995 DM,

reform=1 for the 1998 panel data, 0 otherwise;

The first few lines of drvisits.dat look like

Sabre commands

out drvisits.log

trace drvisits.trace

case obs

yvar numvisit

family p

constant cons

dis m

dis e

dis m

dis e

case id

dis m

dis e

dis m

dis e

stop

Sabre log file

<S> trace drvisits.trace

<S> data id numvisit age educ married badh loginc reform summer obs

2227 observations in dataset

<S> case obs

<S> yvar numvisit

<S> family p

<S> constant cons

Iteration       Log. lik.       Difference

__________________________________________

1          -7502.8826

2          -6041.6484        1461.

3          -5943.5095        98.14

4          -5942.6925       0.8170

5          -5942.6924       0.8498E-04

<S> dis m

X-vars            Y-var

______________________________

cons              numvisit

reform

age

educ

married

summer

Univariate model

Standard Poisson

Number of observations             =    2227

X-var df           =     8

Log likelihood =     -5942.6924     on    2219 residual degrees of freedom

<S> dis e

Parameter              Estimate         Std. Err.

___________________________________________________

cons                  -0.41286          0.26913

reform                -0.14047          0.26580E-01

age                    0.43611E-02      0.13031E-02

educ                  -0.10653E-01      0.60102E-02

married                0.41662E-01      0.27869E-01

summer                 0.10216E-01      0.40409E-01

Initial Homogeneous Fit:

Iteration       Log. lik.       Difference

__________________________________________

1          -7502.8826

2          -6041.6484        1461.

3          -5943.5095        98.14

4          -5942.6925       0.8170

5          -5942.6924       0.8498E-04

Iteration       Log. lik.         Step      End-points     Orthogonality

length    0          1      criterion

________________________________________________________________________

1          -4859.3150        1.0000    fixed  fixed       830.40

2          -4642.0020        1.0000    fixed  fixed       2502.3

3          -4578.7722        1.0000    fixed  fixed       607.18

4          -4560.8207        1.0000    fixed  fixed       289.02

5          -4556.2858        1.0000    fixed  fixed       225.79

6          -4555.0430        1.0000    fixed  fixed       315.35

7          -4552.5998        1.0000    fixed  fixed       13.605

8          -4552.4882        1.0000    fixed  fixed       155.27

9          -4552.4881        1.0000    fixed  fixed       181.54

10          -4552.4881        1.0000    fixed  fixed

<S> dis m

X-vars            Y-var             Case-var

________________________________________________

cons              numvisit          obs

reform

age

educ

married

summer

Univariate model

Standard Poisson

Gaussian random effects

Number of observations             =    2227

Number of cases                    =    2227

X-var df           =     8

Scale df           =     1

Log likelihood =     -4552.4881     on    2218 residual degrees of freedom

<S> dis e

Parameter              Estimate         Std. Err.

___________________________________________________

cons                   -1.2784          0.50028

reform                -0.56688E-01      0.47767E-01

age                    0.51025E-02      0.25158E-02

educ                   0.92880E-02      0.11103E-01

married                0.11339E-01      0.52326E-01

summer                -0.51553E-01      0.75535E-01

scale                  0.90172          0.19394E-01

<S> case id

Iteration       Log. lik.       Difference

__________________________________________

1          -7502.8826

2          -6041.6484        1461.

3          -5943.5095        98.14

4          -5942.6925       0.8170

5          -5942.6924       0.8498E-04

<S> dis m

X-vars            Y-var

______________________________

cons              numvisit

reform

age

educ

married

summer

Univariate model

Standard Poisson

Number of observations             =    2227

X-var df           =     8

Log likelihood =     -5942.6924     on    2219 residual degrees of freedom

<S> dis e

Parameter              Estimate         Std. Err.

___________________________________________________

cons                  -0.41286          0.26913

reform                -0.14047          0.26580E-01

age                    0.43611E-02      0.13031E-02

educ                  -0.10653E-01      0.60102E-02

married                0.41662E-01      0.27869E-01

summer                 0.10216E-01      0.40409E-01

Initial Homogeneous Fit:

Iteration       Log. lik.       Difference

__________________________________________

1          -7502.8826

2          -6041.6484        1461.

3          -5943.5095        98.14

4          -5942.6925       0.8170

5          -5942.6924       0.8498E-04

Iteration       Log. lik.         Step      End-points     Orthogonality

length    0          1      criterion

________________________________________________________________________

1          -4886.1681        1.0000    fixed  fixed       511.63

2          -4718.6770        1.0000    fixed  fixed       3227.1

3          -4665.3282        1.0000    fixed  fixed       2056.0

4          -4655.1747        1.0000    fixed  fixed       706.59

5          -4652.5395        1.0000    fixed  fixed       643.44

6          -4651.1922        1.0000    fixed  fixed       335.76

7          -4650.3290        1.0000    fixed  fixed       213.16

8          -4649.6786        1.0000    fixed  fixed       144.40

9          -4649.0833        1.0000    fixed  fixed       146.90

10          -4648.5772        1.0000    fixed  fixed       51.409

11          -4647.9713        1.0000    fixed  fixed       33.725

12          -4647.6488        1.0000    fixed  fixed       11.365

13          -4647.4903        0.1250    fixed  fixed       15.728

14          -4647.3924        1.0000    fixed  fixed       26.114

15          -4647.3923        1.0000    fixed  fixed       31.637

16          -4647.3923        1.0000    fixed  fixed

<S> dis m

X-vars            Y-var             Case-var

________________________________________________

cons              numvisit          id

reform

age

educ

married

summer

Univariate model

Standard Poisson

Gaussian random effects

Number of observations             =    2227

Number of cases                    =    1518

X-var df           =     8

Scale df           =     1

Log likelihood =     -4647.3923     on    2218 residual degrees of freedom

<S> dis e

Parameter              Estimate         Std. Err.

___________________________________________________

cons                  -0.88590          0.47693

reform                -0.40069E-01      0.31762E-01

age                    0.11508E-01      0.23579E-02

educ                   0.10743E-01      0.13271E-01

married               -0.10293E-01      0.50940E-01