Binary really should be numeric as the summary functions are ever so useful. If your S1 to S6 are numeric then   sum(of s1-s6) = Number of 1s n(of s1-s6) - sum(of s1-s6) = Number of zeros n(of s1-s6) = number of responses   You character version also loses information for any observation where not all of the 6 variables were populated. Quick if you see "1011" which 4 of the 6 had a response? You get the same result for S1=1 S2=0 S3=1 and S4=1 as for S1=1 S2=missing S3=0 S4=1 and S5=1, S1=1 S2=missing S3 = missing S4=0 S5=1 and S6=1 and a few more combinations.   If your s1 to s6 are numeric then you likely had problems with what catx does with the conversion to character as without some additional control you would have blanks in the converted values. ... View more

Many thanks for your reply. This is really helpful and really appreciate your help. ... View more

Hello,   If I look at this paper :   Paper 5500-2016 Latent Class Analysis Using PROC LCA Patricia A. Berglund, University of Michigan https://support.sas.com/resources/papers/proceedings16/5500-2016.pdf , your calling of the macro seems correct (on the condition that it was defined and compiled of course). Use the options  options mprint symbolgen mlogic source2; and run again. Then you get additional log on what may be wrong. Please note PROC LCA itself is a user-defined procedure not officially supported by SAS.   Cheers, Koen ... View more