# Ordinal Independent Variables and Dependent Variable Regression Analysis

I have 3 variables (2 are independent and 1 is dependent). All three variables are measured with questionnaires consisting of multiple questions recorded on 5 point scale. I want to perform regression analysis to check for mediation. How do I go about doing that? Do I have to transform or recode variables? My variables are explained below

Creativity = Q1 + Q2 + Q3 (Do I need to transform Q1-Q3 into single variable representing all three q's of creativity to run regression; if so how to do that?)

Climate = Q1 + Q2 + Q3 + Q4 + Q5 + Q6 + Q7 + Q8 (Q1-Q8 into single variable)

Motivation = Q1 + Q2 + Q3 (Q1-Q3 into single variable)

All the questions (Q1, Q2, ...) are measured on a 5 point scale, 1 for strongly disagree and 5 for strongly agree. I want to run below model.

Creativity $=$ constant $+\ \beta_1$ Climate $+\ \beta_2$ Motivation + error

I am using SPSS.

• Is Q1 for Motivation the same as Q1 for Climate, etc? Or are you independently numbering the questions within each domain? Commented Aug 22, 2016 at 19:04
• Search ordinal independent variables. Commented Dec 9, 2016 at 8:17

1. Likert scale is also known as summated score. You can average the score to get a single variable from different items. Alternatively, you can use Factor Analysis for dimension reduction.
2. For multiple regression, you can simply use your model

Creativity == constant +  β1 Climate +  β2 Motivation + error

3. If there is literature support for presence of mediation effect. You can use Sobel's test (If sample size is good enough) or bootstraping to test for it.

Analysis should come after theory. Only on the basis of past evidences, you can take variable 'motivation [or climate] as a mediator or moderator [More complex relationship may occur], analysis for both is different'

Make sure to test for model assumptions as well.