If significnat level alpha 5% and it is two tails test, P value is 4%. Since P< Alpha , it can be rejected. My question is that do we have to divide P by 2 before compared with alpha becuase it is two tail test?
thanks
Nope.
agreed, no need to divide by 2.
when we compare alpha and p value, no need to care about 1 or 2 tails test, the decision rule is the same 
2 tails test, signicficant level is 5%, p value 4%, compare 5 vs 4 => reject H0
1 tails test, confident interval is 95%, p value 4%, => alpha 5%>p value=> reject H0