https://communities.sas.com/t5/Statistical-Procedures/Re-Chi-square-vs-proc-glimmix/m-p/632061#M30300 <P>The tests are for different questions.</P><P>&nbsp;</P><P>Depending on the type of chi-square test, you are probably only testing if there are significant proportional differences across the eight (2x4) categories simultaneously.&nbsp; So you might get significance, for example, if any 2 combinations of categories is has a smaller N than each of the the other six combinations of categories.</P><P>&nbsp;</P><P>Proc glimmix with the multinomial option will test if your binary predictor is significantly better at predicting each of three of your outcome categories relative to a fourth outcome (reference) category.&nbsp; So here the question is if your binary predictor is associated with a significantly higher or lower probability of being being in one, two, or three of your outcome categories compared to its probability of being in the fourth reference outcome category.</P> Fri, 13 Mar 2020 20:15:10 GMT tellmeaboutityo 2020-03-13T20:15:10Z https://communities.sas.com/t5/Statistical-Procedures/Re-Chi-square-vs-proc-glimmix/m-p/632061#M30300 <P>The tests are for different questions.</P><P>&nbsp;</P><P>Depending on the type of chi-square test, you are probably only testing if there are significant proportional differences across the eight (2x4) categories simultaneously.&nbsp; So you might get significance, for example, if any 2 combinations of categories is has a smaller N than each of the the other six combinations of categories.</P><P>&nbsp;</P><P>Proc glimmix with the multinomial option will test if your binary predictor is significantly better at predicting each of three of your outcome categories relative to a fourth outcome (reference) category.&nbsp; So here the question is if your binary predictor is associated with a significantly higher or lower probability of being being in one, two, or three of your outcome categories compared to its probability of being in the fourth reference outcome category.</P> Fri, 13 Mar 2020 20:15:10 GMT https://communities.sas.com/t5/Statistical-Procedures/Re-Chi-square-vs-proc-glimmix/m-p/632061#M30300 tellmeaboutityo 2020-03-13T20:15:10Z