Before running a binary logistic regression model i was interested to know the strength of association between IV and DV but for some independent variables the results came out to be insignificant.. so would it be wise to carry those variables to the logistic regression ?
Following the recommendations of Hosmer, Lemeshow, and Sturdivant, you may include into your initial multiple logistic regression model IVs with
p-values up to
0.25 on univariable analysis (i.e., analysis using the DV and each IV separately). You also may include IVs of known practical relevance into the initial model regardless of their significance in the univariable analysis. Then, the model can be modified and refined in subsequent iterative steps.