I've done a survey and got results slightly late and running out of time of my dissertation due.
I need some help on the statistical analysis i need to perform?
- Questionnaire test ?
- Inferencial analysis ?
I have a very basic question confusing me :
my survey is measuring satisfaction, it consists several dimentions (5) with each one have 2 questions.
how to get the satisfaction scores?
I can calculate it manually actually by averaging all the likert scores i have collected, does this work ? :(
Ahmad, it is difficult to help without more information, such as Study Objective(s), Study Design (which is the survey you are referring to), Sample Size, Response Variables, Nature of Data collected on each response variable from each respondent, ...
At this point in your process, you should consider working with a statistician in your department or school. You'll want close collaboration to ensure that the work is performed correctly and that you understand the results. I assume you'll need to defend your dissertation in front of a committee? Even more reason to have a local statistician do the work.
I second what Ariel has suggested to you. The Ph.D. committee will check whether you or someone else has done the statistical analysis, so it is important that you maintain high ethics by bringing in a statistician who can defend her/his work, while you, as a Ph.D. candidate, will be responsible for understanding the statistical results.
I agree with Ariel and Raid that you probably need a professional to help you, but make sure that it is someone who is familiar with applied analysis, rather than just someone who teaches statistics. Too often, high level experts propose complex solutions to problems that can be handled more simply.
Meanwhile, because there have been so many questions about analyzing Likert-type items, I have stated a thread here that contains basic resources on this topic:
https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items
There are several options available to you for a suitable statistical analysis for your project. I wish I knew more about your project to be able to advise you better here. It is useful to consider doing more than one analysis.
Example: Collapse the scores for the two highest categories and the two lowest categories and throw out the middle score. Then, make one response "1" (maybe for strong agreement and agreement) and a "0" for no agreement and strong disagreement. Then, run a logistic regression.
Another approach would be to "pretend" that the averaged Likert scale scores are continuous, and then use a method that assumes you have continuous data.
It is a lot of fun to play with data.
It doesn't sound like Ahmed has a lot of time for play, unfortunately.
In fact it would be standard practice in many fields to average the 2 items for each of the 5 scales, although it would have been highly desirable to have more than 2.
One way to check how well the 10 items worked in terms of forming 5 scales would be to print out a correlation matrix that is arranged by the scales. That is, the first 2 items in the correlation matrix would correspond to the first of the 5 scales, then the next 2 items would correspond to the second scale, and so on.
What you are hoping to see are two things. First, there should high correlations whenever two items are from the same scale (each of these will be along the first diagonal). Second, all the other correlations should be lower because they are between items from different scales.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis