Likert Scale Help

#1
Hello everyone!

I am in a statistics course and writing my first research paper using my own data analysis and I could really use some help determining what statistical test to use.

I am using STATA and ANES data. All of my variables are likert scale variables which I have coded -3 to 3 with 0=neither agree nor disagree/neither favor nor oppose. Some of them have -3 as strongly disagree and others have -3 as strongly agree. I did this because some of the questions treat the topic differently than others and I wanted to combine some of my variables into a scale. I also have simplified versions of the variables that are coded -1 to 1. My hypotheses are fairly simple: I expect that an increase in a certain set of IVs will predict an increase in my DV, and that another set of IVs will not correlate with my DV.

I know that from here I will do some tabs with chi-squared and set up some bivariate tables, but beyond that I am completely lost. To make matters worse, this is kind of a crash course in statistics and I feel like a lot of the things I'm seeing we haven't even covered. We have covered tabulations, chi-squared, Cramer's V, tau-b, and the following specific commands: ttest, cor, reg, logit. The only ones I really feel confident with are tab, chi-squared, cor, and reg. The UCLA site seemed to imply that I should use ologit, but I don't know if I should interpret the resulting table the same as a regular logit table, and I found sites that conflict by saying that ologit should only be used for dichotomous variables. I could recode dichotomously but I feel like I would lose a lot of the robustness of my analysis that way. I found a lot of step-by-step this-is-the-stat-test-you-should-use type sites but a lot of them conflicted, asked questions I didn't understand, or said there was no test for this type of data (probably because I answered a question wrong). I have also found a lot of conflicting information about whether or not is appropriate to treat this type of data as interval data.

Ay ay ay! Sorry, a little venting towards the end there. In short, I've been sitting here playing with variables and combing the web all day and when I close my eyes I see STATA and I'm getting nowhere. Any help would be greatly appreciated!!
 

Karabiner

TS Contributor
#2
I expect that an increase in a certain set of IVs will predict an increase in my DV, and that another set of IVs will not correlate with my DV.
So you want to aggregate your set of IVs in an appropriate manner and correlate
the resulting aggregate variable(s) with your DV (I suppose the DV is interval scaled).
You could aggregate the items which belong to the same scale by simply summing them up. You can use Pearson or Spearman correlation then.


Kind regards

K.