I am working on a complex question about how to use summary input data form plus and how to appropriately specify the number of observations. I already checked the mplus user guide and several related threads in the forum, but could not find something that would help me out.
My data comprises thousands of cases who responded to items of a larger test. As answering all items of the test would be too time consuming for the participants, they only responded to a random sub set of all items (for example, 6 items out of the overall pool of 19 items).
Nonetheless, I would love to analyze the factor structure by using CFAs. I therefore thought about building a matrix of pairwise item-intercorrelations and to use this matrix as type = correlations.
With this rather complex pattern of numbers of observations/missings how would I then specify the most appropriate number of observations in Mplus? Should I go for the number of participants who have at least 1 item they responded to? Or would there be another more appropriate approach? I thank you for all your efforts in advance.