Logistic data separation

CathyVI — Thu, 23 Jun 2022 23:10:45 GMT

Hello,

I am trying to use IPTW but I encountered this warning after running my code. I would like to know what this warning means and how to rectify the warning and have a smooth running code.

/*Log*/

1 OPTIONS NONOTES NOSTIMER NOSOURCE NOSYNTAXCHECK;

73 proc logistic data= data1 desc;

74 class sex agerace (ref=first)

75 ALZHDMTE_a (ref=first)

76 AMIE_a (ref=first) ANEMIA_a (ref=first)

77 ATRIALFE_a (ref=first) ASTHMA_a (ref=first)

78 CHFE_a (ref=first) CHRNKDNE_a (ref=first)

79 COPDE_a (ref=first) DIABTESE_a (ref=first)

80 GLAUCMAE_a (ref=first) HIPFRACE_a (ref=first)

81 HYPERL_a (ref=first) HYPERP_a (ref=first)

82 HYPERT_a (ref=first) HYPOTH_a (ref=first)

83 ISCHMCHE_a (ref=first) OSTEOPRE_a (ref=first)

84 RA_OA_E_a (ref=first) STRKTIAE_a (ref=first)

85 cancer_a (ref=first) /param=ref;

86 model case (event='1') = age sex race ALZHDMTE_a

87 AMIE_a ANEMIA_a ATRIALFE_a ASTHMA_a CHFE_a

88 CHRNKDNE_a COPDE_a DIABTESE_a GLAUCMAE_a HIPFRACE_a

89 HYPERL_a HYPERP_a HYPERT_a HYPOTH_a ISCHMCHE_a

90 OSTEOPRE_a RA_OA_E_a STRKTIAE_a cancer_a / link=logit rsquare;

91 output out=data1_PS p=A1_C xbeta=logit_ps;

92 run;

NOTE: PROC LOGISTIC is modeling the probability that case=1.

WARNING: There is possibly a quasi-complete separation of data points. The maximum likelihood estimate may not exist.

WARNING: The LOGISTIC procedure continues in spite of the above warning. Results shown are based on the last maximum likelihood

iteration. Validity of the model fit is questionable.

NOTE: There were 1234567 observations read from the data set WORK.data1.

Re: Inverse probability weighting

ballardw — Thu, 23 Jun 2022 19:45:25 GMT

It is caused by your data. Fix is problematic.

You can read this link for some ideas that may apply to your data: https://stats.oarc.ucla.edu/other/mult-pkg/faq/general/faqwhat-is-complete-or-quasi-complete-separation-in-logisticprobit-regression-and-how-do-we-deal-with-them/

Note that is the first return from my internet search of " quasi-complete separation of data points"

topic Logistic data separation in SAS Programming

Logistic data separation

Re: Inverse probability weighting