which is better/how they differ? independent t-test or simple logistic regression? As a beginner in stats, i am wondering do simple logistic regression do what independent t-test does? and if yes, why do we still do t-test ?
e.g. If there is a binary variable (e.g. smokers to non-smoking )  and a continuous variable (e.g. age)
I can use t-test to test the mean difference between the 2 groups 
I can use simple logistic regression and consider smoking as outcome and age as predictor
I think i can also use bisereal correlation ??
are they different ?
 A: All three procedures can yield, if this is of interest, p-values telling the probability that results as far or farther from the usual null hypothesis would occur if chance alone were responsible.  Beyond that, each procedure will give a different type of potentially useful information:


*

*Point-biserial correlation will yield a coefficient ranging from -1 to 1, 
summarizing (in somewhat abstract or scale-free terms) the degree of
connection between age and smoking status.  (Note that the lesser-used "biserial correlation" works somewhat differently:  see explanation).

*T-testing will
   describe the mean age difference between smokers and non-smokers in
   terms of the standard error of such a difference.

*Logistic
   regression will yield the difference in (the odds of being a        smoker) that is associated with each unit (presumably year) 
   difference in age.  Logistic regression can also be used to generate a probability that each person is a smoker, based on age.
Thus in a given research situation these procedures may well turn out to be complementary.
