If the three independent variables are measured on a nominal scale of measure, you can recode each as dummy variables (0 and 1) and compare all three at one time to predict the dependent variable. If the dependent variable is interval or ratio, use linear regression. If the dependent variable is nominal, use logistic regression.
Thank you very much for the answers. The three independent variables are nominal/categorical and, as Peter said, I can recode them as dummy variables ( 0 and 1); while the dependent variable is continuous. In fact, the data for the dependent variable are collected through a test and thus the data are scores. Should I use linear regression, then?
I agree with Peter's suggestion re: coding them as dummy variables. Depending on the software you will use, you may not need to code as dummy. Most stat software like SAS /STATA/SPSS can handle categorical variables without the need to dummy-code.
As for type of regression, it depends on the distribution of your outcome variable. If it's normally distributed, or can be transformed to be normally distributed, then you can use linear regression.
Thanks a lot for the supplementary explanation. So, I will use linear regression. I will also use SPSS as I haven't worked with the other stat software such as SAS or STATA.