Should I use multiple t-tests?

want to do a statistical test to find out the effect of 3 binary ("yes/no") variables on one continuous variable.





For example, is there a statistical difference in "happiness score, from 1 to 100" between people in the following groups?





Sex M/F


Obese Y/N


Computer-user Y/N





What would be the best statistical test to use?


I thought that 3 t-tests (for effect of Sex, Obesity and then Computer-use) would work.





Would it be easier if there were only 2 yes/no variables (Obesity and use of computer)?





If I use multiple t-tests, do I need a Bonferroni correction (and can someone explain what that really means? I read it in a book).|||Why not use OLS regression? If the dependent variable (happiness score) is continuous, then you can intrepret the b coefficients as the effect on happiness for the one unit increase in Sex, Obese, Computer-user. They would be directly comparable. This would give you the independent effect of each characteristic.





Yes, you could use t-tests, so long as you calculate the mean and standard deviation for each of the three pairs of comparison groups. The disadvantage of this approach is that you would not likley be able to account for the possibility of, say, a greater number of men than women using computers, or a greater numbe of men or women who are obese. Basically, regression does this for you, but three seperate t-tests on 3 pairs of means will not.|||If it's binary, it may be best to use an r x k analysis. Student t-tests are good for continuous data, where as ratios and matrices are good for discrete data.|||No, the t-test is for comparison of two mean values and you must know the sd as well.


In this case you need a 3 x2 khi square test|||Try the below website, it tells you how to adjust your critical values based on the fact that you are doing three t-tests