Between-Groups Analysis Before Logistic Regression?

#1
Hello,

I have a binary outcome variable (categorical) and multiple predictor variables (some continuous, some categorical). I would like to run a binary logistic regression to determine whether the model can predict outcome group. Is it feasible to conduct a between-groups analysis (e.g., one-way ANOVA or chi-square) on each variable prior to the logistic regression to determine if there is a significant difference between the two groups and then only incorporate the significant variables into the model?

Thanks in advance!
 

hlsmith

Omega Contributor
#3
@j58 - correct, this is sometimes called "table 2 fallacy." In addition, you would mis-represent mediators and effects of outcomes, which always get missed if context knowledge isn't incorporated.

Plus ANOVA take three categories.