Dear all,
I am fairly new to SPSS and will have some very basic questions (which look like the most complicated phenomenon to me) about how to approach my data analysis.
I have designed a questionnaire form my research project which is a curriculum evaluation study. I have a number of sections in the questionnaire such as demographic information of the participants (gender, university graduated, years of teaching experience...etc), sections about the different aspects of the curriculum to be evaluated (e.g goals and objectives, course content and materials...etc). Except for the demographic info section, the rest of the items are on a 5 point Likert Scale (SA A N D SD)
The first question is about the reliability check. When I wish to do it on SPSS, should I include the items in the demographic information part or should I add the items on 5 point Likert scale (total 51 items).
The second question is about the descriptive statistics and it is pretty much the ame as the first one. Whether I should include the Likert scale items only or the demographic info should be added too?
That's all for now.
Thanks in advance.
Kind Regards
Ahmet
Hi,
You have to estimate the reliability of your scale only with your 51 items (5 point Likert scale).
But be careful, the reliability estimation is based on Pearson correlations which are adapted for metric data. With ordinal data, you have to use another computations:
- Kappa: https://journals.sagepub.com/doi/abs/10.1191/0962280205sm413oa?journalCode=smma
- or, better for me, polychoric correlations (see attached article).
Best
CH
Summarize your Likert scale data using descriptive statistics. Exercise caution in this step. A common mistake is to calculate a numerical average, or mean value of the coded responses. This is not a valid method for analyzing Likert scale data, which are ordinal in nature.
As an alternative, summarize your Likert scale data with the mode, or the most frequent response. For example, if “agree” was the most frequent response to an item, the mode would be the numerical value assigned to that response.
Explore the data further with inferential statistical techniques. Many such techniques exist, and the most appropriate one will depend on the exact nature of your study. Analysis of variance is one approach. For the example in Step 1, you could analyze responses with the respondent’s gender as an independent variable, examining the difference in responses between male and female survey participants. Factor analysis, which tries to explain responses as a function of underlying factors, is also used frequently.
Finally, there is some disagreement in education and research about whether you should run parametric tests like the t-test or non-parametric hypothesis tests like the Mann-Whitney on Likert-scale data. Winter and Dodou(2010) researched this issue, with the following results:
"[T]he t test and [Mann-Whitney] generally have equivalent power, except for skewed, peaked, or multimodal distributions for which strong power differences between the two tests occurred. The Type I error rate of both methods was never more than 3% above the nominal rate of 5%, even not when sample sizes were highly unequal.”
Hope some of this helps.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS