I am working on new research to identify the reliability of the different sets of questionnaires corresponding to different learning styles. I collected and cleaned the data and loaded it into the SPSS software. I have data for around 600 learners from different backgrounds of learning. For questionnaire set 1, I have only 38 variables that measure the questions' uncertainty and use values 0 for uncertain, 1 for false, and 2 for true. I am not getting the results I want. The total factors should be 5 but I am getting around 14 whose eigen values are 1 or greater than 1 in the software. What am I doing wrong here? Please help.
Your coding scheme is strange. Why is the uncertain category not between the false and true categories?
In general, eigenvalues suffer from the problem that there is no definition of an important eigenvalue. If you are expecting five subscales, you can examine this directly using Mokken scaling, which identifies item subsets that meet criteria for a unidimensional scale.
If you think there are five, and your observed data are categorical (how many uncertain responses were there), you might consider a confirmatory nominal response model. But you need to provide some details about why you doing this (e.g., are you trying to summarize the data in fewer variables, do you have a particular model for how these five theoretical variables influence each other, do you want to use these to differentiate groups). Also, you need to provide code to show what you did if you want people to try to identify what you did wrong.
Hello Ankita Jiyani Mangtani. Like other respondents, I am not really sure what you want to do. But I can focus on one aspect of what you wrote:
The total factors should be 5 but I am getting around 14 whose eigen values are 1 or greater than 1 in the software.
The eigenvalues > 1 rule of thumb is not recommended these days, as it tends to produce models with too many factors. See the following article for a nice overview of that issue and several others that may be relevant for you.
- https://kuscholarworks.ku.edu/bitstream/handle/1808/1492/preacher_maccallum_2003.pdf
HTH.
Ensure that you have appropriately prepared your data, selected suitable factor analysis settings, and interpreted the results in a meaningful way. Factor analysis can be a complex technique, and it is essential to validate and corroborate the results with existing literature or theoretical frameworks.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey