Hello guys,
I have done some Google research regarding my topic and he brought me here. I actually found some similar threat, but i think they are way too much for my 101 class.
I have to deal with Likert Scale type variables, they are subordinate variables for my main independent variables that i need to analyze using basic methods only: ttest, one-way anova, chi-squared test, single/multi regression and principals component analysis (which I think can be used for regression)
My dependent variable is test score, one of my main independent variables is parental influence, which consist of Mother, Father level of education (categorical) and Likert-scale interaction variables (how often help doing home work, how often discuss school. etc...). 1 variable is 3-scale as follows:
Array
Others are in 4 scales as follows:
Array
After searching around Google and friends, I can only think of runny one-way anova for mother/ farther education separately and then running pca for "interaction" variables to get new "score" variables to be included in the regression model.
My 1st question is: Is it okey to do so? Is there a better way which is suitable for my 101 Class?

My second question is also related to Likert Scale variables. My second main independent variable is "Student attitude toward learning". There are various Liker Scale variables in the dataset, using same scale, one example is followings:
Array
I also have in mind to run pca for all the variables to find main component variables to be included in regression(1). However, i also try to combine all of them under "Attitude" category by using " egen newv1 = rowmean (var1 var2 var3 ...) as suggested by someone else, so that i can create new "attitude" variable to be included in my regression model (2). However, result of (1) and (2) is totally different in terms of p value. Then, why is it so?

Thank you very much,