Hi, I'm doing a research on quality requirements of an industry.
My questionnaire (likert scale type) has 4 parts, each part is made up of several questions and directly relates to one factor/requirement. I have done the Croabach reliability test to confirm the relationship of all questions to its respective components.
The thing is when i run the data as a whole chunk some questions from factor 1 seem to relate to factor 2, whereas others seem to relate better with Factor 3. And I'm also having issues with multilinearity. The thing is what I'm looking for is the relationship between the 4 components. Is it possible to create composite variables for each dimension to see the correlation? I'm pretty new to SPSS so I'm not really sure how to go about it.
Hi Sha,
some points: first of all do a Croabach analysis separately for every variable with its respective measures (the questions designed for every variable). What score do you get for every variable? then conduct a exploratory factor analysis (EFA) to understand to what extent each measure is related to its respective variable and in summery to see how strong is your variables classification. Further step: what is the model in which you are testing the relationships among variables? structural model, linear model? the method through which you have to analyze the relationships depends on the model.
Dear Sha,
I think firstly, you prepare about pilot study. because, you will go to fill you questionnaire in industry directly, there will many questions will not suitable for your research so better way firstly you will have to prepare pilot study questionnaire and go in industry fill it and then you will come to make final questionnaire.
The initial solution under EFA will give an output with several components and the questions (variables) will seem to cross load on two or more components. Given that your questionnaire has four parts, i suspect you must be looking for four factors. I will suggest that after the initial solution out put, run the EFA again and under the extraction command, instead of using extract based on Eigenvalues, choose extract based on fixed number of factors and define 4 factors. Under the options command, tick sorted by size and suppress small coefficients (define less than 0.5). The variables should clearly load on four components (factors).
Dear Sha,
I suggest Andy Field's book, Discovering Statistics using IBM SPSS Statistics.
You can also see the help of SPSS. In the menu Help, choose Topics, then search for Factor Analysis. If you choose the second option you can see how to perform an EFA, namely a PCA.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS