Actually, it is not appropriate to treat ordinal data as if it were continuous interval data. Think about it this way: how much more is "definitely agree" than "agree"? And is the "definitely agree" the same distance from "agree" as "agree" is from "disagree"?

In regards to presenting the data, I actually like the presentation of ordinal data in a bar chart. It's straight forward in its interpretation and easy to visualize.

I agree with Ariel,  although there is possibility to treat ordinal variables as scales, without/ by avoiding the narrative explanation  (1-not agree at all, 2 -not agree etc) but only as values/grades (from 1 to 5, from 1 to 10 etc). 

Here are some useful links and a paper (attached):

http://www.statisticssolutions.com/data-levels-and-measurement/

http://gerardnico.com/wiki/data_mining/measurement

It really depends. A likertscale has 7 points, and 'disagree' is not necescarily less or more than 'agree'. However, in social sciences, it is not likely that the question you asked is a single-item measurement. In fact, most of the research I've ever seen use multi-items measurments, in which each item is added and then divided by the amount of items the scale has. In that way a 7-item likertscale with 5 items usually generates more than 30 possible outcomes (although it potentially can reach up to thousands of possible outcomes in this example).

Therefore, the likertscale is no longer a true likert scale, but instead a semi-continuous measurement which should be treated as such, that is, as an interval variable.

Hope this helps

As suggested in an earlier comment, it is generally (professionally) acceptable follow convention with your discipline.

Mathematically, ordinal measures, such as a likert scale, do not offer sufficient variation (and often do not offer anything near a 'normal' distribution) to use parametric statistics.

However, if your intention is, as Sander notes, to convert these items into a scale/index which follows the form of an I/R measure, we bypass the mathematical complications.

Just as importantly (or more, depending on who you ask), there is a theoretical issue here. If we understand a respondent's choice as a reflection of some underlying continuous concept (e.g. 'depression' is more than just the sum of 'agrees', but a continuum of affect) then latent variable analyses may be the best way to present the models. A good conceptualization of this is outlined in:

Lee, S.-Y., Poon, W. Y., & Bentler, P. M. (1995). "A two-stage estimation of structural equation models with continuous and polytomous variables". British Journal of Mathematical and Statistical Psychology, 48, 339–358.

I agree with Marija. We recommend using 7 point Likert questions with only the end numbers anchored to interpretations. We believe 7 is ideal due to working memory models giving the maximum richness of data balanced against the ability of the human brain to discriminate 7 individual options.

In terms of descriptives, I would recommend using Excel rather than SPSS to create bar charts. If you are comparing several groups then I'd suggest creating a percentage frequency bar chart. This can be done using pivot tables. You can then import your raw data into SPSS, changing any text values into numbers.

To be precise, a Likert scale is in fact a group of questions with Likert responses. If these questions have been shown to be all about the same thing then the responses can be added together to create a single scale variable. This even works with traditional 5 point Likert questions where every value is anchored to an interpretation. This process is often known as validation and requires a reliability analysis with a sample size of at least 300.

Undermeasurement is a very old and still ongoing discussion, which as Alex mentioned depends very much on your discipline and your research topics. From a mathematical point view it is not acceptable.

But from a pragmatic point of view it often makes sense, alltough you have to accept some uncertain biases.

Perhaps it might help you to read these classics in this context:

Stevens, Stanley S. (1946): On the Theory of Scales of Measurement. In:

Science 103, 1946 (2684), S. 677–680.

Jaccard, James/Wan, Choi K. (1996): LISREL approaches to interaction

effects in multiple regression. Thousand Oaks, Calif.

In my opinion, ordinal data seems to have a meaningful differences as appose to likert scale. Depends on what you are trying to interpret.

Yes it really depends.  I also came across some researchers treat ordinal data (e.g. Likert Scale) as interval data by modifying their survey questionnaire i.e. converting Likert scale questionnaire (e.g. 1=Effective, 2=Some How Effective, 3=Moderately Effective, 4=Neutral, 5=Moderately Not Effective, 6=Some How Not Effective, 7=Not Effective) to Semantic Differential scale using Bipolar Adjectives (e.g. Effective1, 2, 3, 4, 5, 6, 7 Not Effective) so that parametric tests can be applied. The rationale claimed is that the intervals between the scale values can be treated as equal intervals, making it an interval scale which justified for parametric tests.  What's your view on this?

Weighted mean is another valid option

I believe the ordinal data should not be treated as continuous, as the interval between strongly disagree to disagree, or disagree to neutral may not be equal.