Suppose your dependent variable (DV) is a Likert scale or something similar. That is, it’s some sort of rating, from 1 to 5 or 1 to 7 or some such. And suppose you want to regress that on several independent variables. What should you do?

There are three broad categories of regression models that might be applicable. A lot of people routinely use linear regression (often simply called regression). Others routinely say this is incorrect, and that you should use ordinal logistic regression. And yet others will do things such as multinomial logistic regression, or collapsing the DV into two categories, and then doing binary logistic. Which is right?

The short answer to this is to quote Sir David Cox

There are no routine statistical questions, only questionable statistical routines

Let’s get more specific. Suppose you are a doctor studying back pain, and suppose your DV is response to a scale:

How much pain are you in on a typical day

1 – None

2 – Barely noticeable

3 – Moderate

4 – Severe

5 – Excruciating

and your independent variables are things like age, sex, injury status, time since injury and so on.

If one is strict about it, linear regression requires a continuous DV – and we do not have one, at least as we’ve measured it, although it could be argued that there is a latent underlying variable here that is continuous. But you’d be hard pressed to prove that the difference between “none” and “barely noticeable”

is the same as that between (say) “moderate” and “severe”. Technically, if you follow Steven’s categories of nominal, ordinal, interval, ratio, your DV is ordinal, and should be analyzed with some form of ordinal logistic regression.

But the most common type (by far) of ordinal logistic regression is the proportional hazards model, which assumes proportional hazards. That assumption might be violated, in which case, you might want to use multinomial logistic.

Since those are relatively unusual methods, some people just collapse the categories into (say) “severe'” or “excruciating” vs. anything less than that.

Which is right?

The great advantages of linear regression are its ease of interpretation and its familiarity. But it might be wrong.

Ordinal logistic is more likely to be correct, but is less known and harder to understand.

Multinomial logistic is even harder to understand, and is a very complex model, with many parameters to estimate.

Collapsing the variable will only very rarely be correct. It throws away information, and that’s rarely a good thing to do.

So, here’s what I recommend:

Do ordinal logistic regression and test the assumptions. Then if the assumptions are met, also do linear regression and compare the results by making a scatterplot of one set of predicted values vs. the other. If they are very similar (YOU decide. Statistical analysis requires thought and judgment) then go with linear regression. If the assumptions are NOT met, then also do multinomial logistic regression, and compare those two sets of results, opting for the simpler ordinal model if results are very similar.

I specialize in helping graduate students and researchers in psychology, education, economics and the social sciences with all aspects of statistical analysis. Many new and relatively uncommon statistical techniques are available, and these may widen the field of hypotheses you can investigate. Graphical techniques are often misapplied, but, done correctly, they can summarize a great deal of information in a single figure. ** I can help with writing papers, writing grant applications, and doing analysis for grants and research.**

** Specialties:** Regression, logistic regression, cluster analysis, statistical graphics, quantile regression.

You can **click here to email** or reach me via phone at 917-488-7176. Or if you want you can follow me on Facebook, **Twitter**, or LinkedIn.

I am not an SPSS user but probably factor is for categorical variable and covariate for continuous. But that’s a guess

I’m having a hard time interpreting the ordinal regression. Is it okay for me to run linear regression instead?

No, linear regression is probably not going to be OK. The assumptions will almost surely be violated, plus it assumes that the scale of the DV is interval.

Thank you for the fast reply,

Can someone please explain what is Test of Parallel lines, Model Fitting Information and Goodness-of-fit in basic human language?

I have little to no background in statistic. All I wanted to know is what IV have impact on the DV.

I cannot teach all of ordinal logistic regression here. If you would like to hire me to help you with your analysis, let me know.

I am doing my dissertation, and now struggling with SPSS, my lecture told me to use ordinal regression because both of my variables,Dependent and Independent, are categorical (Likert-scale rating). Both have the same scale because they are asking the same questions, because I want to know if for instance Brand Association has a significant positive impact on Brand Equity. So what do I need to know. How can I come to my conclusion if my indpendent variables have an impact on my dependent variable? Please help me I am trying to understand regression since two days, still did’t get it.

Hi Merve

Regression usually takes an entire semester, and then logistic regression (which is where ordinal logistic comes in) another semester.

If you’d like to hire me to help with your data analysis, let me know, but explaining all of regression is not something I can do in a blog post.

Peter

I have a challenge in analyzing this type of data from likert use scale, from 200 respondents I collected with total number of 41 questions. I want to rank it and come out with the severity or more critical hypothesis.

THE MAJOR CHALLENGES IN OUR REFINERY IS ATRRIBUTED TO ONE OF THE FOLLOWINGS SA

5 A

4 UND

3 D

2 SD

1

1. Government policy

2. Lack of funding

3. Lack of personnel training

4. Political intervention and hindrances

5. Frequent changes of GMD

6. Corruption in the system

7. Dependence on product importation

8. Lack of adequate facilities to drive the production

9. Lack of personnel incentive and motivation

10. Lack of board management’s good decision

11. Shading deal of NNPC officials

12. The cabal involvement in product lifting

13. Activities of militant in the Niger Delta

14. Activities of oil theft among Niger Delta

15. vandalization of pipe line in the nation’s pipeline net work

16. Present insurgency of Boko Haram against genuine investors

17. Insecurity in the entire Niger Delta

18. Inadequate experienced man power to manage the refinery

19. Tax concession

20. Inadequate number of refinery (since we have just four)

21. Lack of proper deregulation of the down- stream sector

22. Inconsistence government policy

23. Delay in the implementation of major project in the refinery

24. Inability to make effective decision as to allocation of resources

25. Inability to manage and control resources from the federal government

26. Inability of passing the yearly budget for quick implementation

27. Awarding of contracts to unqualified and incompetent contractors based on personal influence

28. Lack of monitoring project and investigation of abandoned project.

29. Graft and shared of resources meant for project

30. Lack of adequate and routine turn around maintenance

31. Environmental oil pollution

32. Community settlements and its activities

33. Lack of focus on the part NNPC to play its statutory roles in the nation’s oil sector.

34. Lack of proper auditing from the office of the Auditor general.

35. Lack of PIB`s passage and implementation

36. Lack of adequate maintenance of equipment such as: pump, compressor, turbine heat exchanger Boiler and others

37. Most of the refinery’s facilities and equipment are obsolete

38. Inability to make effective decision as to allocation of resources

39. Graft and shared of resources meant for project

40. Inability of passing the yearly budget for quick implementation

41. Awarding of contracts to unqualified and incompetent contractors based on personal influence

Hi, I don’t understand what you want to do. What do you mean by “come out with severity or more critical hypothesis”?

You come up with your hypotheses

beforeyou do the analysis, not after.Dear Sir,

I have completed research for training and development.

My independent variables are content of the training program, trainers ability, trainees capacity and dependent variable is employee performance.

I have included 07 questions for 1st variable, 06 questions for 2nd variable and 04 questions for 3rd variable. All questions are in likert scale of Strongly Disagree, Disagree, Neutral, Agree, Strongly Agree.

Please help me how to conduct correlation and regression test in order to test hypothesis.

Your guidance in this regard would be highly appreciated.

Thanks,

Sampath.

This sounds like you first need to do factor analysis on each of the sets of variables that make up the independent variables. Then you need to do some kind of regression, which kind will depd on how employee performance is measured.

Hello, I\’m currently doing a dissertation on Consumption. I\’ve been asked to use OLS, Ordered Probit and Ordered Logit. I have already conducted the required tests. But now the hardest part is the analysis, should I insert separate output tables for each? Thanks

I don’t understand. The tests are part of the analysis.

You certainly need separate tables for each type of regression.

hi, i have an interval dependent variable, can i run a multiple regression? Please i would appreciate your response.

Yes. At least, the fact that the DV is interval is not a hindrance to multiple regression. It’s a requirement.

Hello Sir, have a problem with analysing my data because my IV is in likerty scale and so is my DV. I understand i can use an ordered logistic regression but is that well suited given the nature of my IV. thanks.

Hello Sir, have a problem with analysing my data because my IV is in likerty scale and so is my DV. I understand i can use an ordered logistic regression but is that well suited given the nature of my IV. thanks.

It is a good method because of the nature of your DV, as you supposed. There are no really good commonly used methods for Likert scale variables but you could look into optimal scoring if you are interested in less used approaches.

Hello I have developed a survey that consists of:

Dependent Variables:

• 26 items with Likert scale (strongly agree, agree, disagree, strongly disagree and don’t know)

• 16 items with Likert scale (Very helpful, somewhat helpful, not very helpful, not helpful at all and don’t know)

Independent Variables:

• 6 categorical, 2 continuous and 6 dichotomous

QUESTION: Because the dependent variables (survey items) all have ordinal responses, I should use Ordinal logistic regression, correct?

Thank you in advance for any guidance.

If you have 42 different dependent variables, then something is very odd about your study. If you are adding the Likert items to make a new scale then you are already assuming that they are interval, so you can use regular regression.

Hi

How do you calculate values for dependent variable if you only have results for various independent variables obtained from Lickert Scale?

Hi Rachael

>>>How do you calculate values for dependent variable if you only have results for various independent variables obtained from Lickert Scale?

First, you run a regression. There are no common methods for ordinal independent variables, but you can treat them as categorical or interval. If you are into other methods, you can try optimal scaling. Then you use the parameter estimates from the regression to score new cases.

Peter

Hi, I’m developing a interactive tool for my final year project which takes Likert scale questions related to population satisfaction with their local authority, with ratings from 1-6 including Neutral and Don’t know responses and try to predict the level of satisfaction in the future. The data frame contains only 4 years of data, but the prediction would have to be up to 2050 (according to the population projection growth of another data frame). What method should I use, since I have Likert scale questions as my DV and only the different years as my IV? Many thanks for your help!

This really isn’t possible to do.

Hi,

I have a likert scale for my DV which asks the respondents how likely they would be tempted to purchase a beer brand based on an advert. I have two types of adverts that are shown and therefore was planning on conducting two linear regressions and comparing attitudes towards the two.

I have 5 IVs which are gender, age education, income and current frequency of purchase.

Is multiple linear regression the right form of analysis to be conducting for this study?

Thanks

I have a likert scale for my DV – (How likely would you be tempted to purchase this beer brand based on this advert)

My IVs include – Gender, age, education, income, and current frequency of purchase of beer.

Is linear regression the correct for of analysis to use in this case?

No, you probably want ordinal logistic regression, but, if the proportional odds assumption is violated, then probably multinomial logistic

hi, i have data with likert scale which measures to what extent nurses miss essential care: 1=never ,2=rarely, 3= ocasionaly, 4= frequently and 5= always. I want to look at the effect of my IV(AGE, educational status, year of experiance, type of hospital they work etc on DVwhich is likert) can i utilize linear regresion taking mean of dv?

thank you

No. You want ordinal logistic regression.

Hi, I have data responses in the form of a 10 point likert scale (four groups, randomized to differing interventions). I believe I can compare responses (liker scale response) between groups using a Kruskal Wallis test.

However, I am also asking such questions such as age, gender, socioeconomic status etc to each group. In order to understand if the aforementioned variables have an effect on my likert scale response, is ordinal logistic regression the correct method? Would I undergo ordinal logistic regression on each group?

Yes, ordinal logistic is right and no, you would do it on the groups together with group being one of the independent variables.

Sir, I have 3 independent variable with 5 likert scale in each variable and dependent variable with 4 likert scale. so i was wondering how to fit my data with a regression model.

Would really appreciate your help

Sir, I have 3 independent variable with 5 likert scale in each variable and dependent variable with 4 likert scale. so i was wondering how to fit my data with a regression model.

Would really appreciate your help

Hi Prabin

If your DV is the sum of 4 likert scales, then probably you can use ordinary regression.

Peter

Sir, would you please clarify what you meant:

In second paragraph from the bottom: “So, here’s what I recommend:

Do ordinal logistic regression and test the assumptions. Then if the assumptions are met, also do linear and regression and compare…” What do you mean by ‘do linear and regression’?

It’s a typo; the word “and” should be deleted. I will do that now.

Hello,Sir I am doing my academic research using likert scale including always,often,sometimes,rarely,never and having difficult to find the base for merging these items on discussion.

Please I would appreciate your response. Thank you

I don’t understand the question. What base? There’s no base that I know of. Combining Likert items is tricky. Factor analysis may be the best way.

Hi

I have a 5 point Likert Questionnaire on Total Quality Management with “Totally disagree,Disagree,Neutral,Agree and Totally agree.I have 18 questions about TQM implementation in organizations to which respondents reply as per the Likert scale.I would like to know how to test this as the 18 questions relate to 6 research questions.

Many thanks

There’s no way to answer without knowing what your research questions are, how the 18 questions relate to each other and so on. You can send me e-mail at peterflomconsulting@mindspring.com