Variable selection in Logistic Regression

Hello all,
I have a query regarding attrition analysis using logistic regression. Say that I calculate the length of stay of an employee in an organization (say 'Tenure') and divide tenure in to 10 buckets . Every employee falls in to any one of these buckets(1 to 10). Say percentage attrition is relatively more in first three buckets(1,2,3) and last two buckets(9,10) relative to the other ones(4,5,6,7,8). i.e. the graph of attrition percentage over the buckets decreases and then increases(U- curve).
As there is no monotonous increase or decrease in the attrition percentage across the buckets, should I be discarding this variable from the model?

Thanks a lot in advance!


Less is more. Stay pure. Stay poor.
Why dont you ditch the buckets and try inserting tenure in as a quadratic term or if you are savvy a spline term.