Is it valid to transform a 5 point likert scale question [Strongly agree-Strongly disagree] to a four point scale by combing strongly disagree and disagree [due to low frequency in both strongly disagree and disagree] or would I have to combine strongly agree and agree as well.
Thanks.
Combining it to a four point scale may give your data different meaning entirely. Adding up "strongly Disagree" and "disagree" may even be more than the response for "Agree" thereby countering the interpretation you ought to have. So, i dont think the idea of combining scale is the best option.
Thanks for the response. I understand the concerns regarding interpretation. However, i'm planning on using the variable in a LOGIT model and and when the frequency count from some categories are below 5, it is my understanding that will create issue in the model.
You may also try and testing for outliers and see if that removes participants who have contributed to the low frequency on the question. You may end up skewing your result if you combine the categories. Something that may not have been statistically significant, may now be so because you combined categories to increase their frequency. There is a trade off. I would run the model multiple times with and without combining the categories and see if it makes a difference in outcomes.
What do you want to do with the data?
If you want to make predictions, you can use something like a Random Forest. I had a project where I needed to look for drivers of bad survey results. Traditional methods didn't work well. RF worked wonders.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis