1

I have 2 independent variables with 2 categories each: 'gender' (male, female) and 'relationship status' (married, single), and one DV: 'emotional impact' measured on a 10 point scale ranging from 1 (least) to 10 (most). What test do I use? Would it be ordinal logistic regression?

User1865345
  • 8,202
lisaarthur
  • 21

1 Answers1

2

Assuming that your outcome is a discrete, ordinal measure, the textbook recommendation would be to use ordered logistic regression. However, you will find that scholars sometimes use OLS instead of ordered logistic regression. They might provide various justifications for this decision (higher number of response categories, ease of interpretation, etc.). You can estimate both models, compare the results, and run some diagnostic tests to understand possible violations of assumptions and how sensitive your analysis might be to modeling decisions. However, you should not use these analyses to arrive at estimates that best suit your interests.

T.E.G.
  • 2,332
  • Thank you for your kind response. My DV is a scale from 1 to 10, 1 being the least and 10 being the most, the variable is emotional impact. I've never heard of ordered logistic regression. Could you recommend me a textbook which would have an explanation of it? Classic textbooks don't even include ordinal logistic regression. – lisaarthur Oct 19 '23 at 13:18
  • 1
    @lisaarthur You are welcome. The terms ordered and ordinal are used interchangeably. Proportional odds model is also used. There are various online sources, e.g., this OARC tutorial. This one is in R, but they also have tutorials in Stata and SPSS. You can find chapters on ordered/ordinal logistic regression in many books, e.g., Agresti's Analysis of Ordinal Categorical Data or Harrell's Regression Modeling Strategies. A more introductory text is O'Connell's Logistic Regression Models for Ordinal Response Variables. – T.E.G. Oct 19 '23 at 15:19
  • I will have a look, thank you. Can I do an ordinal logistic regression where one of the independent variables is also ordinal (a scale from 1 to 10)? I can't find such examples online. – lisaarthur Oct 19 '23 at 16:18
  • 1
    @lisaarthur Yes, you can. Predictor specification is similar to other regression models. This question might also be of interest to you: https://stats.stackexchange.com/questions/195246/how-to-handle-ordinal-categorical-variable-as-independent-variable – T.E.G. Oct 19 '23 at 16:39
  • I have just been told to treat my ordinal DV as a continuous variable. Do you know how to do it? I think they know I want to avoid doing ordinal logistic regression because even after reading up on it, I still don't get it. It's not easy like ANOVA or multiple regression. And I cannot find it in the Andy Field book, he does not cover it at all. – lisaarthur Oct 19 '23 at 17:40
  • 1
    You can treat your ordinal variable as continuous and run OLS (I don't know which software you use but it's a simple step). Keep in mind that this might not be a better strategy. Field's book is an introductory text. It is not surprising that it does not cover some modeling approaches. – T.E.G. Oct 19 '23 at 17:54
  • I'm watching an OLS tutorial and they are explaining the parameter estimates table. It is completely insane. They say: "Age (one of the IVs) has a plus sign which means it is falling into the higher category on the dependent variable, so in this case, it is higher interest." - How do I translate this to my situation? So if males have a plus sign there, they are falling into the higher category of my DV, which is emotional impact.? So could I say males were falling under number 10 on my scale, and were more emotionally impacted? This is so not specific. – lisaarthur Oct 19 '23 at 18:05
  • I only ever use SPSS. But Andy Field covers everything else! – lisaarthur Oct 19 '23 at 18:06
  • 1
    People usually translate their findings into statements such as "controlling for gender, married respondents have higher scores on emotional impact compared to single respondents" and they report effect sizes to give an idea how much higher, etc. But it feels like you are trying to run before you walk. Start with simple analyses and try to understand how they extend to more complicated ones. Unfortunately, these comments are now beyond the scope of the original question, so I will stop here. – T.E.G. Oct 19 '23 at 18:26
  • 1
    I understand. Thank you so much, you have helped immensely. – lisaarthur Oct 20 '23 at 11:51
  • Sorry to be bothering you again. I have watched a couple of tutorials on OLS and it is making sense but then it gets to a thing called odds ratios. SPSS does not calculate this and I am not familiar with other sofware. And secondly, I cannot find a template on how to report OLS results (APA style). Any advice? – lisaarthur Oct 22 '23 at 20:58
  • 1
    Do you mean ordinal logistic regression? OLS does not report odds ratios. – T.E.G. Oct 22 '23 at 23:42
  • Yes, ordinal logistic regression, not OLS! I have just checked now what OLS means and have never come across such. – lisaarthur Oct 23 '23 at 13:00
  • 1
    Ordinary least squares (OLS) is an estimation method commonly used in multiple linear regression. If you have run a linear regression model, chances are you have already used OLS. Any tutorial explaining how to report multiple linear regression results in APA format would work. – T.E.G. Oct 23 '23 at 15:31