- For 5 point Likert scale questions and binary (0 or 1) dependent variable, which regression analysis should I use? Can I use a binary logistic regression on SPSS?
- There are two or three questions to measure one independent variable. I run a factor analysis after that two questions are tied together. Now should I make another column in SPSS data sheet to find average value of both tied questions? And use that average value in regression analysis? If do average then average will come into points for example, 2.5, 3.5 etc. Is this points coding OK to run regression analysis?
My answer is based on making some assumptions about your question.
Assumption 1: Dependent variable is binary.
Assumption 2: Independent variable (predictor) is ordinal.
If these assumptions are accurate, then the answer is yes. A binary logistic regression would be appropriate as the type of regression one runs is dependent on the dependent variable. A helpful hint is that if possible, have your independent variables have a meaningful 0 (e.g., "never" or "neutral" or sample mean). This tends to help with interpretation of the regression weights.
Based on the information you provided, it sounds as if you have the individual items of a scale (or subscale). For example, if I have 3 questions about different symptoms of a disease, they should factor together. I could average these items or sum the items to create a scale score that is then used as an independent or dependent variable in regression. This is commonly done in many different fields (particularly any field that uses surveys).
Hope this is helpful.