I am looking for a proper method for my research. I want to analyze left-right political position of a person.

My idea is to combine logit regression with a linear regression. Logit will decide on weather a person is on left or right side of the spectrum: it will consist of factors like attitude towards immigrants or toleration towards minorities.

The other part will define the extremity of person's views. It would consist of factors like being satisified with current government or other factors that push people towards extreme political views.

I am looking for anything that can somehow help me find a proper tool. Thanks!

  • $\begingroup$ Do you have left-right labels, or is this part supposed to be unsupervised? $\endgroup$
    – Tim
    Nov 27, 2018 at 21:56
  • $\begingroup$ Its a variable from 0(left) to 10(right). Its based on questionary data. $\endgroup$
    – jadas08
    Nov 27, 2018 at 22:08

1 Answer 1


What I understand from your question is that you have some kind of questionnaire data including political views on scale 0 - leftish to 10 - rightish and some additional, explanatory variables. For me this is a regression problem, where given the explanatory variables you want to predict the degree of leftism vs rightism. You would probably want to re-code your data to -5 (leftish) to +5 (rightish) for better interpretability (sign would be political orientation, while magnitude its extremity). Since the outcome variable is bounded, you rather should not use linear regression in here, but something designed for such cases, e.g. beta regression.

The nice part about such model is that it preserves the continuity of the left-right dimension, that seems to have been assumed by someone who designed the questionnaire. Under above model, different factors can influence how someone "moves" over the dimension and this seems to be consistent with the above assumption.

  • $\begingroup$ But what about the part which covers only extremity od views? It should push the views towards 0 or 10 but we dont knows which way. As i understand the beta regression dosent cover that part $\endgroup$
    – jadas08
    Nov 28, 2018 at 8:35
  • $\begingroup$ @JędrzejAdaszyński what kind of conclusions would you like to make based on this data? Could you give an example of the conclusion you would like to draw? The interpretation for most cases would be the same, you would just flip, or ignore the signs of the parameters. $\endgroup$
    – Tim
    Nov 28, 2018 at 8:51
  • $\begingroup$ For example being unsatisfied with government pushes a person towards extreme views - either extremely right or extremely left. The influnce of government satisfaction wont necessarily make a right wing extremist, but propably will make the views more extreme $\endgroup$
    – jadas08
    Nov 28, 2018 at 10:26
  • $\begingroup$ @JędrzejAdaszyński OK, I get your point, now this makes more sense. Maybe you could edit your question to add the example? $\endgroup$
    – Tim
    Nov 28, 2018 at 10:47
  • $\begingroup$ @JędrzejAdaszyński you can check stats.stackexchange.com/questions/320924/… and stats.stackexchange.com/questions/81457/… and cran.r-project.org/web/packages/flexmix/vignettes/… for examples of some models that you can find interesting $\endgroup$
    – Tim
    Nov 28, 2018 at 10:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.