2

I am curious to know whether it is current practice in the statistical field to interpret and report data obtained on a Likert Scale (1: Strongly Agree to 5: Strongly Disagree), by finding the median value rather than averaging all the values for a particular item. Please see link here

My second question is how do we decide whether we should use 1: for Strongly Disagree rather than Strongly Agree such that: 1: Strongly Disagree 2:Disagree 3:undecided 4:Agree 5:Strongly Agree

1: Strongly Agree 2:Agree 3:undecided 4:Disagree 5:Strongly Disagree

user39531
  • 587
  • 3
  • 8
  • 17
  • Which end is 1 and which end is 5 is completely arbitrary. The conclusions should be the same. Select whatever goes with what's done in your field / literature. – John May 29 '14 at 04:35
  • Are you asking about estimating central tendency from your sample by taking the mean or median of responses to a single item from all respondents? @John: which end is 1 may be arbitrary on the part of the person who designs the scale, but it may affect conclusions. See [my answer here](http://stats.stackexchange.com/a/96880/32036). – Nick Stauner May 30 '14 at 00:50
  • If we do an average of the number of responses x strongly agree (1) +...+number of responses x strong disagree (5) and conversely as above, we will get two different mean value. That is why I would like to know how do we decide whether we should use 1: for strongly disagree or 1: for strongly agree – user39531 May 30 '14 at 01:04
  • Nick, there are differences in how they're presented to subjects if they see the numbers, that's true. This is about coding them post hoc...or so I thought. If it is about how it's presented to people doing the rating then my comment should be ignored. – John May 30 '14 at 01:17
  • Indeed, I am more interested to know how do we attribute the coding 1-5, to each Likert Type, post-data collection. – user39531 May 30 '14 at 01:43
  • If you've already collected the data, how do you not already know whether 1 refers to strong agreement or disagreement? – Nick Stauner May 30 '14 at 23:08
  • 1
    Dear user39531, you asked 21 questions on this site and got for 16 at least one answer. But your vote count shows only ONE up vote and nothing else. Moreover, no answer has ever been accepted by you. Accepted answers and up votes are, however, sort of the daily meal for people offering their time in order to help people which only contribute by questions - people like you. Please appreciate answers with votes, accept them and make us smile. – random_guy Jan 10 '15 at 19:09

1 Answers1

1

If you just have the one scale neither is current practice in the sense that it's the most common or best advice. It's probably better to use median and it would have to be a pretty special case to recommend mean. You should probably look up ordinal regression, why it's used, and how to interpret it. Sometimes it helps to look at R or SPSS examples.

John
  • 21,167
  • 9
  • 48
  • 84