Hello,
I have seen more questions being asked about this topic. I have Likert scale data and some outliers. Not extremely many, but there are outliers. I've read that outliers do not really exist when it comes to Likert scale data. I have decided to keep the outliers in my research and not delete them, because looking at the data, they are all realistic answers.
I was wondering if anyone had found any literature on this, so that I can justify/substantiate leaving the outliers in my data.
Everything depends on what you mean by outliers. Let's assume the dependent variable ls a likert variable and thus ordinal. I would therefore recommend ordinal logistic regression as the analysis method . Since your DV is ordinal there really can't be outliers on this side. Same for the likert variables on the RHS of your model. If you have any other IVs in your model there could be x-outliers. I would recommend plots of non-likert variables to try to identify them. TBC
I am running a multiple regression analysis with all Likert-scale variables. I have one dependent variable and 9 independent variables. The sample size is big, 1205.
sorry, I'm late returning. See these links:
Reference thesis for outliers in likert scale data - Cross Validated (stackexchange.com)
and
https://www.google.com/search?q=X+outliers+in+linear+models+in+regression&rlz=1C1CHBF_enUS917US917&oq=X+outliers+in+linear+models+in+regression&aqs=chrome..69i57j33i22i29i30.29129j0j15&sourceid=chrome&ie=UTF-8
BTW x-outliers are also called leverage points by some authors
These references talk about identifying leverage points. Another reference is:
https://www.researchgate.net/post/Is_it_always_appropriate_to_categorize_a_continous_variable_in_to_a_categorical_variable_and_consider_logistic_regression_analysis
and ordinal logistic regression
https://www.google.com/search?rlz=1C1CHBF_enUS917US917&sxsrf=ALeKk02dL0mYV6iLuIROcBhimc4GegWGew%3A1609363185642&ei=8e7sX4fUJtLdtAbHmrCoBw&q=ordinal+logistic+regression+likert+variables&oq=ordinal+logistic+regression+likert+variables&gs_lcp=CgZwc3ktYWIQAzIFCAAQzQIyBQgAEM0CMgUIABDNAjIFCAAQzQIyBQgAEM0COgQIABBHOgcIABDJAxBDOgQIABBDOgIIADoFCAAQyQM6BggAEBYQHjoFCCEQoAE6BQghEKsCOgcIIRAKEKABULBmWM6jAWC-qQFoAHACeACAAXOIAaIMkgEEMTMuNJgBAKABAaoBB2d3cy13aXrIAQjAAQE&sclient=psy-ab&ved=0ahUKEwjHkPqH0fbtAhXSLs0KHUcNDHUQ4dUDCA0&uact=5
and one more:
https://www.researchgate.net/post/How_do_I_identify_outliers_in_Likert-scale
Best Wishes, David Booth
Elisa de Lange , actually I believe you do not need to justify your decision of leaving the outliers. The justification of a decision of leaving the outliers depends on what are outliers in the first place. I would not expect outliers in a Likert scale because the scales are not designed to allow for outliers, and hence I would found awkward to even talk about outliers in the context of your research.
- Badges
- Science topic
- Similar topics
- Automotive Engineering
- Machines