SPSS Statistics

 View Only
Expand all | Collapse all

Arbitrary covariance matrix possible in mixed?

  • 1.  Arbitrary covariance matrix possible in mixed?

    Posted 20 days ago

    Hi All,

    I'm curious if there is any way to directly specify the random slopes and intercepts covariance matrix for MIXED?

    Specifically, I have a factor A with two levels, plus a pre covariate, with a post outcome. The factor A and covariate pre are clustered within another factor B, which also has two levels, and we want to treat the intercepts and sloples for the A|B and A*Pre|B as random.

    Dealing with each of these separately would require a 2 x 2 diagonal covariance, and dealing with them simultaneously (if I have understood everything) is then two 2 x 2 block diagonal matrices, with a possible correlation between the 2 slopes and the two intercepts ... I don't see any obvious choice for a covariance structure from those available in SPSS v29. Have I simply missed something or how would one specify such a covariance structure?

    Thanks,
    Colm.



    ------------------------------
    Colm McGuinness
    ------------------------------