I have a closed questionnaire to access student's knowledge of a topic. Each questions have several (3, 4 or 5 generally) options. In each question there are one true answer and several not true. However I feel I can order them from the far from true answer, thus assigning for instance a 1, to the true answer at which I assign the higher figure in that question, e,,g, 5. I make the assumption, valid to me, that some answers have more true content than others.
Can I the apply descriptive statistics as if I am working with quantitative variables?
Hi Paulo
Your transformed variables will be basically ordinal in nature. Just be aware that descriptive statistics such as mean and standard deviation, should be used with caution, especially with variables with less than 5 categories. You would be better off giving frequency tables and graphs.
Hi, Paulo,
Adrian is correct about individual items; however, if the items produce a summative overall value, then the total scale should be treated as interval data and analyzed quantitatively.
And note that even ordinal data can be analyzed with the proper statistical tests. Spearman rho for correlations. Mann-Whitney U test for two group comparisons.
The biggest challenge for you is defending your ranking of the answers. I would suggest locating a collaborator to also rank the answers so that you can report inter-rater reliability in assessment of the answers.
Adrian and Peter
Thanks to both of you for the most helpful answers, suggestions and hints. Just a note: way that caution related to less than five items. For instance, if score the items as I said (and secured a well and strong justification for that score and not other) and in a three item responses I have a mean of, say, 1.2000 and a std of 0.1000, can't I say with a great confidence that all respondent's didn't get the write answer and so they rely heavily on the misconception corresponding to the item assigned with the score 1?
Thanks
I keep learning on this topic from the answers given to you by my fellow professors. Applied statistics is fascinating.
Being myself a production management professor, I devote a considerably proportion of my time to devising, testing and applying data gathering questionnaires for field research work. This is why, before examining any figures, as such, we try to make sure that the set of data being collected truly and fully depict the situation under analysis.
I would also like to point out that the weighting of qualitative (or semi-qualitative) answers does not usually follow a straight line, which we usually adopt for ease of assessment. Consequently, we should keep in mind that this simplification entails a departure (¿small?) from the observed reality. This consideration will help us reach some valid conclusions when the set of data that we have collected is quantitavely or semi-quantitativel examined. Best of luck.
- Badges
- Science topic
- Similar topics
- Education