topic Re: A question about PROC SURVEY logistic? in Statistical Procedures https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490876#M25494 <P>Are there any messages in the LOG regarding convergence?</P> <P>&nbsp;</P> <P>Typically with coefficient estimates that large and correspondingly small standard errors, there are convergence issues.&nbsp; It seems like as well that your weights are extremely large.&nbsp; Have you tried normalizing them?</P> Wed, 29 Aug 2018 15:09:15 GMT SAS_Rob 2018-08-29T15:09:15Z A question about PROC SURVEY logistic? https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490868#M25493 <P>Hi,&nbsp;</P><P>&nbsp;</P><P>I'm trying to run a proc&nbsp;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.&nbsp;</P><P>&nbsp;</P><P>The SURVEYLOGISTIC Procedure</P><TABLE><TBODY><TR><TD><P><STRONG>Model Information</STRONG></P></TD></TR><TR><TD><P><STRONG>Data Set</STRONG></P></TD><TD><P>WORK.N15</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Response Variable</STRONG></P></TD><TD><P>Antibiotic</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Number of Response Levels</STRONG></P></TD><TD><P>2</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Stratum Variable</STRONG></P></TD><TD><P>VARSTR</P></TD><TD><P>VARIANCE ESTIMATION STRATUM - 2015</P></TD></TR><TR><TD><P><STRONG>Number of Strata</STRONG></P></TD><TD><P>165</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Cluster Variable</STRONG></P></TD><TD><P>VARPSU</P></TD><TD><P>VARIANCE ESTIMATION PSU - 2015</P></TD></TR><TR><TD><P><STRONG>Number of Clusters</STRONG></P></TD><TD><P>349</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Weight Variable</STRONG></P></TD><TD><P>PERWT</P></TD><TD><P>FINAL PERSON WEIGHT, 2015</P></TD></TR><TR><TD><P><STRONG>Model</STRONG></P></TD><TD><P>Binary Logit</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Optimization Technique</STRONG></P></TD><TD><P>Fisher's Scoring</P></TD><TD><P>&nbsp;</P></TD></TR><TR><TD><P><STRONG>Variance Adjustment</STRONG></P></TD><TD><P>Degrees of Freedom (DF)</P></TD><TD><P>&nbsp;</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Variance Estimation</STRONG></P></TD></TR><TR><TD><P><STRONG>Method</STRONG></P></TD><TD><P>Taylor Series</P></TD></TR><TR><TD><P><STRONG>Missing Levels</STRONG></P></TD><TD><P>Included (MISSING)</P></TD></TR><TR><TD><P><STRONG>Variance Adjustment</STRONG></P></TD><TD><P>Degrees of Freedom (DF)</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Number of Observations Read</STRONG></P></TD><TD><P>23363</P></TD></TR><TR><TD><P><STRONG>Number of Observations Used</STRONG></P></TD><TD><P>23028</P></TD></TR><TR><TD><P><STRONG>Sum of Weights Read</STRONG></P></TD><TD><P>2.5812E8</P></TD></TR><TR><TD><P><STRONG>Sum of Weights Used</STRONG></P></TD><TD><P>2.5812E8</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Response Profile</STRONG></P></TD></TR><TR><TD><P><STRONG>Ordered<BR />Value</STRONG></P></TD><TD><P><STRONG>Antibiotic</STRONG></P></TD><TD><P><STRONG>Total<BR />Frequency</STRONG></P></TD><TD><P><STRONG>Total<BR />Weight</STRONG></P></TD></TR><TR><TD><P><STRONG>1</STRONG></P></TD><TD><P>No</P></TD><TD><P>21546</P></TD><TD><P>239733598</P></TD></TR><TR><TD><P><STRONG>2</STRONG></P></TD><TD><P>Yes</P></TD><TD><P>1482</P></TD><TD><P>18390178</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Model Fit Statistics</STRONG></P></TD></TR><TR><TD><P><STRONG>Criterion</STRONG></P></TD><TD><P><STRONG>Intercept Only</STRONG></P></TD><TD><P><STRONG>Intercept and<BR />Covariates</STRONG></P></TD></TR><TR><TD><P><STRONG>AIC</STRONG></P></TD><TD><P>132597682</P></TD><TD><P>131164140</P></TD></TR><TR><TD><P><STRONG>SC</STRONG></P></TD><TD><P>132597690</P></TD><TD><P>131164156</P></TD></TR><TR><TD><P><STRONG>-2 Log L</STRONG></P></TD><TD><P>132597680</P></TD><TD><P>131164136</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Testing Global Null Hypothesis: BETA=0</STRONG></P></TD></TR><TR><TD><P><STRONG>Test</STRONG></P></TD><TD><P><STRONG>Chi-Square</STRONG></P></TD><TD><P><STRONG>DF</STRONG></P></TD><TD><P><STRONG>Pr&nbsp;&gt;&nbsp;ChiSq</STRONG></P></TD></TR><TR><TD><P><STRONG>Likelihood Ratio</STRONG></P></TD><TD><P>1433543.99</P></TD><TD><P>1</P></TD><TD><P>&lt;.0001</P></TD></TR><TR><TD><P><STRONG>Score</STRONG></P></TD><TD><P>757520.937</P></TD><TD><P>1</P></TD><TD><P>&lt;.0001</P></TD></TR><TR><TD><P><STRONG>Wald</STRONG></P></TD><TD><P>71303.0062</P></TD><TD><P>1</P></TD><TD><P>&lt;.0001</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Analysis of Maximum Likelihood Estimates</STRONG></P></TD></TR><TR><TD><P><STRONG>Parameter</STRONG></P></TD><TD><P><STRONG>DF</STRONG></P></TD><TD><P><STRONG>Estimate</STRONG></P></TD><TD><P><STRONG>Standard<BR />Error</STRONG></P></TD><TD><P><STRONG>Wald<BR />Chi-Square</STRONG></P></TD><TD><P><STRONG>Pr&nbsp;&gt;&nbsp;ChiSq</STRONG></P></TD></TR><TR><TD><P><STRONG>Intercept</STRONG></P></TD><TD><P>1</P></TD><TD><P>-30.8434</P></TD><TD><P>0.0911</P></TD><TD><P>114655.453</P></TD><TD><P>&lt;.0001</P></TD></TR><TR><TD><P><STRONG>PPI</STRONG></P></TD><TD><P>1</P></TD><TD><P>14.1581</P></TD><TD><P>0.0530</P></TD><TD><P>71303.0062</P></TD><TD><P>&lt;.0001</P></TD></TR></TBODY></TABLE><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Odds Ratio Estimates</STRONG></P></TD></TR><TR><TD><P><STRONG>Effect</STRONG></P></TD><TD><P><STRONG>Point Estimate</STRONG></P></TD><TD><P><STRONG>95% Wald<BR />Confidence Limits</STRONG></P></TD></TR><TR><TD><P>PPI</P></TD><TD><P><FONT color="#FF0000">&gt;999.999</FONT></P></TD><TD><P><FONT color="#FF0000">&gt;999.999</FONT></P></TD><TD><P><FONT color="#FF0000">&gt;999.999</FONT></P></TD></TR></TBODY></TABLE><P>&nbsp;</P><P>&nbsp;</P><TABLE><TBODY><TR><TD><P><STRONG>Association of Predicted Probabilities and Observed<BR />Responses</STRONG></P></TD></TR><TR><TD><P><STRONG>Percent Concordant</STRONG></P></TD><TD><P>4.0</P></TD><TD><P><STRONG>Somers' D</STRONG></P></TD><TD><P>0.040</P></TD></TR><TR><TD><P><STRONG>Percent Discordant</STRONG></P></TD><TD><P>0.0</P></TD><TD><P><STRONG>Gamma</STRONG></P></TD><TD><P>1.000</P></TD></TR><TR><TD><P><STRONG>Percent Tied</STRONG></P></TD><TD><P>96.0</P></TD><TD><P><STRONG>Tau-a</STRONG></P></TD><TD><P>0.005</P></TD></TR><TR><TD><P><STRONG>Pairs</STRONG></P></TD><TD><P>31931172</P></TD><TD><P><STRONG>c</STRONG></P></TD><TD><P>0.520</P></TD></TR></TBODY></TABLE> Wed, 29 Aug 2018 14:42:57 GMT https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490868#M25493 alotaibifm 2018-08-29T14:42:57Z Re: A question about PROC SURVEY logistic? https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490876#M25494 <P>Are there any messages in the LOG regarding convergence?</P> <P>&nbsp;</P> <P>Typically with coefficient estimates that large and correspondingly small standard errors, there are convergence issues.&nbsp; It seems like as well that your weights are extremely large.&nbsp; Have you tried normalizing them?</P> Wed, 29 Aug 2018 15:09:15 GMT https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490876#M25494 SAS_Rob 2018-08-29T15:09:15Z Re: A question about PROC SURVEY logistic? https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490880#M25495 <P>Here is the Log description. No, I haven't normalized the weight.&nbsp;</P><P>&nbsp;</P><P><BR />NOTE: PROC SURVEYLOGISTIC is modeling the probability that Antibiotic='Yes'.<BR />NOTE: Convergence criterion (GCONV=1E-8) satisfied.<BR />NOTE: Only one cluster in a stratum. The estimate of variance will omit this stratum.<BR />NOTE: Only one cluster in a stratum. The estimate of variance will omit this stratum.<BR />NOTE: PROCEDURE SURVEYLOGISTIC used (Total process time):<BR />real time 0.20 seconds<BR />cpu time 0.14 seconds</P><P>&nbsp;</P><P>&nbsp;</P> Wed, 29 Aug 2018 15:13:46 GMT https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490880#M25495 alotaibifm 2018-08-29T15:13:46Z Re: A question about PROC SURVEY logistic? https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490889#M25496 <P>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.</P> Wed, 29 Aug 2018 15:24:41 GMT https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490889#M25496 SAS_Rob 2018-08-29T15:24:41Z Re: A question about PROC SURVEY logistic? https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490892#M25497 <P>Do you have any article or thread about data normalization? I haven't done this procedure&nbsp;before. I'm sorry if this is a basic question.&nbsp;</P><P>I will try to do that and see what will happen.&nbsp;</P> Wed, 29 Aug 2018 15:33:17 GMT https://communities.sas.com/t5/Statistical-Procedures/A-question-about-PROC-SURVEY-logistic/m-p/490892#M25497 alotaibifm 2018-08-29T15:33:17Z