I have number of Likert scale variables in my questionnaire, each scale variable consists of 4 statements. I calculated the median and IQR for each Likert scale variable.
How to interpret the median. For example, one of the variables measuring the usefulness, and contains 4 statements evaluated as (strongly disagree … strongly agree). The median for usefulness is equal to 3.75. how can I explain the usefulness median? Does 3.75 means that the participants agree about the usefulness or what ?
If SA is coded 1, A coded 2, Neutral coded coded 3 Agree coded 4 and SA coded 5 then the median of 3.75 means Agree,
The median score (response) is therefore AGREE.
Best of luck!
Thank you. but what is the general rule to interpret the median in relation to likert scale variable? then I can interpret the other variables I have in my questionnaire.
On a normal distribution curve, the Median is onhesame line with the mean and the mode meaning that all are AGREE.
But if the distribution is positively skewed it means the median is to the right side of the mode but above the mean response which means AGREE is not the mode(lesser) and not the mean (more than the mean).
It it is negatively skewed it means the median (AGREE) is at the left side of the mode, lesser and above the mean.
For Positive skewness, this means that AGRRE is above average (any 1 scale that is the mean) in counts (population) but lower in score.
So therefore,
This means that's your agree could be subject to the skewness.
I am having difficulty understanding your question. Can you provide, for the usefulness variable, what the 4 statements are, how many cases are in each response category, and what numerical scores you assigned to each response category?
I am providing the participants a conceptual framework with three links between game design and learning theories and one of Likert scale variables is usefulness and it includes the following statements:
The first link is understandable
Graphical representations (modelling ELT stages as internal economy) are explicit
The second link is understandable
The third link is understandable
1= strongly disagree, 2= disagree ,……. 5= strongly agree
Mariia Ezhova , the answer is "yes and no." It is quite common to combine (either by summing, or by averaging) several Likert-type questions to form an index. The index is more nearly a continuous, interval-level measure than any single question, and therefore only a statistical purist would object to using the index as either the dependent variable or a predictor variable in a multiple regression model.
Using a single Likert-type question as a predictor in a multiple regression model is sometimes done, but is hard to defend. It would be better to form dummy (dichotomous) variables for the categories of the question, and to enter the set of dummy variables into a regression model in a single step, testing the R-square increase from that step for significance with an F test. That way you are not assuming the question is an interval-level measure, nor that its effect is linear.
Using a single Liker-type question as the dependent variable in a linear regression is ill-advised. In that case, you should probably use ordinal logistic regression instead.
Amani Alsaqqaf , thanks for clarifying. It sounds like you have taken the mean of 4 usefulness questions, each scored 1 to 5, to form a usefulness index scored 1.0 to 5.0. That makes sense, and is quite commonly done with Likert-type questions. The possible scores on the index would be 1.0, 1.25 ( = [1+1+1+2]/4), 1.5, 1.75 ... 4.5, 4.75, 5.0.
Therefore, the median of 3.75 is interpretable just like any other median: "The score for the case in the middle of the distribution of the usefulness index is 3.75. The typical opinion, using the median, is slightly less favorable about usefulness than responses of 'Agree' on all four questions making up the index (which would give a score of 4.0 on the index). It is substantially less favorable than 'Strongly Agree' on all four questions (an index score of 5.0), but more favorable than a 'Neutral' response on all four (an index score of 3.0)."
The median says nothing about "agreement" concerning usefulness. The amount of agreement would be reflected in a measure of dispersion, like the IQR (or the standard deviation), not in a measure of central tendency like the median (or the mean). Low dispersion indicates high agreement.
Because you have taken the mean of the 4 items to form an index, it is not obvious to me why you are using the median and the IQR to describe the distribution of that index rather than using the mean and standard deviation. A multi-item index is more nearly a continuous, interval-level measure than any of the items on which it is based. Having averaged the items to form the index, why not take the average (mean) of the index to describe its central tendency?
Using the mean would allow for much simpler wording of the interpretation, because the mean is more familiar and needs less explanation than the median. Suppose the mean is, say, 3.8. Then the interpretation could be simply: "The average (mean) score on the usefulness index is 3.8. This reflects a slightly less favorable opinion about usefulness than 'Agree' responses on the four questions making up the index, substantially less favorable than four "Strongly Agree' responses and more favorable than four 'Neutral' responses."
I hope this is helpful.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS