Can someone help me out with setting up a statistical model for my analysis?

I'm helping a friend out with the stats portion of her thesis. She has have two continuous independent variables and one dependant variable. The variables are composite scores from a Likert scale questionnaire.





There are two class variables (age and education) which are dichotomous (above or below median).





She;s interested in determining the strength and direction of the relationship between the dep and indep variables. Also she'd like to know how this relationship is affected by whether a respondent is above or below the median age and level of education.





Which statistical analyis is appropriate here? She was instructed that MANOVA is appropriate. I'm unfamiliar with MANOVA and thought that the normal ANOVA analysis would suffice with a class statement (I'm using SAS). Any thoughts or advice would be much appreciated.

Can someone help me out with setting up a statistical model for my analysis?
If it were just two independent and one dependent it would be classic ANOVA. MANOVA is traditionally used for multiple dependent variables. Imagine that you were ran the ANOVA for the entire sample, then run separate ANOVAs for each of the following: above age above edu, above age below edu, below age above edu, and below age below edu. In the separate runs the data would not be symmetrical (unless the age and edu distributions are completely homogenous) so comparing the relationships would be useless. The same applies to if you ran an ANOVA for above age all edu, because if age and edu were at all correlated, the mean edu would differ from the full sample and you couldn't compare the two. In SAS there are corrections for skewed data which you could feasibly use. MANOVA is probably an option too, if you use some sort of repeated measures MANOVA (where one set of independents measures the dependent variable is the other set is age and edu (dichotomous) and they measure the dependent variables of the relationships of the first study) .
Reply:Assuming that a simple model is being used (linear, logistic regression, etc.), you might examine the coefficients assigned to each independent variable, and do the same for models built on just cases below the median age, and another built on just cases above the median age.

Wallpapers software