1
$\begingroup$

I have recently advised some colleagues on the malpractice of binning a continuous variable, which was used in order to put it as a covariate in a regression model and retained as a significant predictor. They responded that when summed responses to several Likert item are not normal (in this case extremely skewed) it is questionable whether these can be used as continuous variables. For this reason, they treated this variable as ordinal data with dummy categories. Is it ok to do so? Isn't it better to transform the variable in order "normalize" it (with methods like square root or loglinear transformations, or aggregating the less frequently observed responses)? Doesn't the binned variables lead to several shortcomings (biased estimates etc.)? Thanks in advance for your answers

$\endgroup$
  • $\begingroup$ Are you talking about binning a single Likert item or the sum of several Likert items? $\endgroup$ – Peter Flom - Reinstate Monica Jun 19 '14 at 10:34
  • $\begingroup$ i am talking about the sum of several likert items. $\endgroup$ – user43897 Jun 19 '14 at 10:55
  • $\begingroup$ OK, then, except in rare circumstances, I'd say binning is not needed. It does depend a bit on what the histogram of the total looks like but, in general, I'd say treating it as continuous is better. $\endgroup$ – Peter Flom - Reinstate Monica Jun 19 '14 at 13:26
  • $\begingroup$ Why would you need to bin to include it as a covariate? $\endgroup$ – Glen_b -Reinstate Monica Jun 19 '14 at 19:46
  • $\begingroup$ I don't know, that's what i am asking them. They suggested that i should read Carifio J, Perla R. Ten common misunderstandings, misconceptions, persistent myths and urban legends about Likert scales and likert response formats and their antidotes. Journal of social science (2007). I've read it but could not find an answer. $\endgroup$ – user43897 Jun 20 '14 at 6:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.