Hi,
I'm trying to run a proc SURVEYlogistic for a complex data, but the output looks weird. anyone has an idea what do those numbers mean? Especially the one highlighted with red.
The SURVEYLOGISTIC Procedure
Model Information  
Data Set  WORK.N15 

Response Variable  Antibiotic 

Number of Response Levels  2 

Stratum Variable  VARSTR  VARIANCE ESTIMATION STRATUM  2015 
Number of Strata  165 

Cluster Variable  VARPSU  VARIANCE ESTIMATION PSU  2015 
Number of Clusters  349 

Weight Variable  PERWT  FINAL PERSON WEIGHT, 2015 
Model  Binary Logit 

Optimization Technique  Fisher's Scoring 

Variance Adjustment  Degrees of Freedom (DF) 

Variance Estimation  
Method  Taylor Series 
Missing Levels  Included (MISSING) 
Variance Adjustment  Degrees of Freedom (DF) 
Number of Observations Read  23363 
Number of Observations Used  23028 
Sum of Weights Read  2.5812E8 
Sum of Weights Used  2.5812E8 
Response Profile  
Ordered  Antibiotic  Total  Total 
1  No  21546  239733598 
2  Yes  1482  18390178 
Model Fit Statistics  
Criterion  Intercept Only  Intercept and 
AIC  132597682  131164140 
SC  132597690  131164156 
2 Log L  132597680  131164136 
Testing Global Null Hypothesis: BETA=0  
Test  ChiSquare  DF  Pr > ChiSq 
Likelihood Ratio  1433543.99  1  <.0001 
Score  757520.937  1  <.0001 
Wald  71303.0062  1  <.0001 
Analysis of Maximum Likelihood Estimates  
Parameter  DF  Estimate  Standard  Wald  Pr > ChiSq 
Intercept  1  30.8434  0.0911  114655.453  <.0001 
PPI  1  14.1581  0.0530  71303.0062  <.0001 
Odds Ratio Estimates  
Effect  Point Estimate  95% Wald  
PPI  >999.999  >999.999  >999.999 
Association of Predicted Probabilities and Observed  
Percent Concordant  4.0  Somers' D  0.040 
Percent Discordant  0.0  Gamma  1.000 
Percent Tied  96.0  Taua  0.005 
Pairs  31931172  c  0.520 
Are there any messages in the LOG regarding convergence?
Typically with coefficient estimates that large and correspondingly small standard errors, there are convergence issues. It seems like as well that your weights are extremely large. Have you tried normalizing them?
Here is the Log description. No, I haven't normalized the weight.
NOTE: PROC SURVEYLOGISTIC is modeling the probability that Antibiotic='Yes'.
NOTE: Convergence criterion (GCONV=1E8) satisfied.
NOTE: Only one cluster in a stratum. The estimate of variance will omit this stratum.
NOTE: Only one cluster in a stratum. The estimate of variance will omit this stratum.
NOTE: PROCEDURE SURVEYLOGISTIC used (Total process time):
real time 0.20 seconds
cpu time 0.14 seconds
Given the large disparity between the sample size and the sum of the weights, you might try normalizing the weights and see if the results make more sense.
Do you have any article or thread about data normalization? I haven't done this procedure before. I'm sorry if this is a basic question.
I will try to do that and see what will happen.
Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.
Register today!ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.