Solved – How to interpret results if a reference category of a categorical variable in multivariable logistic regression is not significant

binomial distributionlogisticmultiple regressionregression

I am trying to do a multivariable logistic regression and using a normal binomial logistic regression, using binomial variable X (coded 0/1) and Y (4 sub-categories). Unfortunately, the reference category was shown to be not significant. Should I then reject the null hypothesis for this variable?

Cat.,   Sig.,   Exp(B)
Y,         1,   .161
Y(1),   .032,   .226
Y(2),   .106,   .370
Y(3),   .061,   .309

Best Answer

As an aside, people usually use Y for the dependent variable and X for the independent variable. It's not necessary to do so, but switching them will confuse people.

What are "subcategories" of Y? For this to be correct, they should be levels of a variable.

What parameterization did you use? (Dummy coding? Effect coding? Something else?)

Next, the reference category is never significant. If you are using dummy coding it is what you are comparing the other levels to. Thus, your results show that Y(1) is significantly different from Y(0), Y(2) is not, and Y(3) is borderline.

Related Question