Model 1

DATA a; INPUT trial treat eclam n; CARDS;

1 1 14 131

1 2 14 136

2 1 21 385

2 2 17 134

5 1 12 1011

5 2 35 760

6 1 138 1370

6 2 175 1336

7 1 15 506

7 2 20 524

8 16 108

PROC GENMOD DATA=a; CLASS trial treat;

MODEL eclam/n=treat trial treat*trial / DIST=B TYPE3 WALD; ESTIMATE 'overall' treat 1 -1 / EXP;

The output is not shown but has a similar form to that from Model 2 below. Model 2

PROC GENMOD DATA=a; CLASS trial treat; MODEL eclam/n=treat trial/DIST=B TYPE3 WALD; ESTIMATE 'overall' treat 1 -1 / EXP;

Model Information

Data Set Distribution Link Function Response Variable (Events) Response Variable (Trials)

Number of Observations Read

Number of Observations Used

Number of Events

Number of Trials

Class Level Information

Class Levels Values trial 9 123456789

treat 2 12

Parameter Information

Parameter

Effect

trial

treat

Prm1

Intercept

Prm2

treat

1

Prm3

treat

2

Prm4

trial

1

Prm5

trial

2

Prm6

trial

3

Prm7

trial

4

Prm8

trial

5

Prm9

trial

6

Prm10

trial

7

Prm11

trial

8

Prm12

trial

9

WORK.A Binomial Logit eclam n

18 18 636 6942

Criteria For Assessing Goodness Of Fit

Criterion

DF

Value

Value/DF

Deviance

8

29

.3761

3.6720

Scaled Deviance

8

29

.3761

3.6720

Pearson Chi-Square

8

28

.7996

3.6000

Scaled Pearson X2

8

28

.7996

3.6000

Log Likelihood

-1877

.3795

The deviance and Pearson chi-square are measures of model fit and have similar roles to the residual sum of squares in normal data models.

Algorithm converged.

Analysis Of Parameter Estimates

Standard Wald 95% Chi-

Parameter

DF

Estimate

Error

Confidence

Limits

Square Pr

>

ChiSq

Intercept

1

-0.

1137

0

.1379

-0.

3841

0.

1566

0.

68

0

.4096

treat

1

1

-0.

4104

0

.0885

-0.

5840

-0.

2369

21.

48

<

.0001

treat

2

0

0.

0000

0

.0000

0.

0000

0.

0000

trial

1

1

-1.

8457

0

.2380

-2.

3122

-1.

3792

60.

14

<

.0001

trial

2

1

-2.

1344

0

.2118

-2.

5496

-1.

7192

101.

52

<

.0001

trial

3

1

-0.

2363

0

.2409

-0.

7084

0.

2359

0.

96

0

.3267

trial

4

1

-0.

5054

0

.2779

-1.

0500

0.

0393

3.

31

0

.0690

trial

5

1

-3.

2740

0

.1958

-3.

6577

-2.

8903

279.

67

<

.0001

trial

6

1

-1.

7287

0

.1420

-2.

0071

-1.

4503

148.

15

<

.0001

trial

7

1

-3.

0515

0

.2150

-3.

4730

-2.

6300

201.

36

<

.0001

trial

8

1

-2.

9293

0

.3830

-3.

6800

-2.

1787

58.

50

<

.0001

trial

9

0

0.

0000

0

.0000

0.

0000

0.

0000

Scale

0

1.

0000

0

.0000

1.

0000

1.

0000

NOTE: The scale parameter was held fixed.

Wald Statistics For Type 3 Analysis Chi-

Source DF Square Pr > ChiSq treat 1 21.48 <.0001

trial 8 444.03 <.0001

Asymptotic Wald chi-squared tests are performed for each fixed effects parameter. These tests should be interpreted cautiously in small datasets.

Contrast Estimate Results

Standard Chi-

Label Estimate Error Alpha Confidence Limits Square Pr > ChiSq overall -0.4104 0.0885 0.05 -0.5840-0.2369 21.48 <.0001

Exp(overall) 0.6634 0.0587 0.05 0.5577 0.7891

The final line of the above output provides the estimate of the OR and accompanying 95% confidence limits, and is generated by the EXP option in the ESTIMATE statement.

Was this article helpful?

0 0

Post a comment