[Rd] about constructing a logit model

[ben]

I am a new B for R, so as this mailing list too.   Thank you very much to those who will reply me and the contrubutors, THANK YOU.
let me tell my problem that I am encountering now
I have grabbed some data from census department to do an analysis.  however, it is too bad that the data have been handle so I cannot treat them as the homework i did in school. 
I am going to test the determinants of the people want to move to developing city from developed city

EXAMPLE suppose there is 3 individuals a,b,c 
a = age <30, salary 10k, and edu.level=highschool    move
b = age 30~40, salary 30k, and edu.level=college      not move
c = age 40~50, salary 20k, and edu.level=highschool   move
(IT IS JUST AN EXAMPLE TO TELL MY SITUATION, I HAVE THOUSAND OF DATA SETS INTEAD OF 3)
I used
<summary(glm(mydata))
and all the variable are significant, but once i do it as 3 models
<summary(glm(move~edu.level))
<summary(glm(move~salary))
<summary(glm(move~age))
only salaries is significant, all the rest pvalue>0.2..too bad
furthermore, i tried
<summary(glm(mydata, family=binomial))
only salary significant too... why does it happen?
the result affects by the "family=binomial"... i m not sure how it works....

further more
actually, my data is not that good as shows as individual..instead...
aged <30  move , 30~40 not move , 40~50  move
 salary.20k move, 30k not move, 40k  move,
edu.level highschool move, highschool move, college  not move

all the categories have been rearranged and i dont know which of each characteristic is with respect to which person.
do u think I will miss some correlation for those factors?
or i can test them group by group only..(move vs age, SAY AGE <30 ARE MORE MOBILITY)
Thank you very much again.
sbest regards
Ben
	[[alternative HTML version deleted]]