Attachment | Size |
---|---|
Moderator.txt | 13.4 KB |
Model.png | 47.86 KB |
Hi Mike and colleagues,
I am trying to estimate a TSSEM model and then test the moderating effect. I am following the wonderful resources provided by you and your team. However, when I compare the constrained and unconstrained models, the difference in degrees of freedom is higher than expected.
In my analysis (model figure and R code attached), I am constraining the A matrix to be equal i.e., I am placing equality constraints on my regression coefficients. My S matrix has 5 parameters - these parameters are not constrained across two models. So, shouldn't my delta df should be 5? However, I get a delta df of 9.
In a subsequent moderator analysis, I am constraining a single regression coefficient to be equal across the two groups. However, my delta df turns out to be 2. I am not sure where I am going wrong. Your help will be appreciated.
Thanks in advance for your help.
Regards,
Srikanth Parameswaran
Hi Srikanth,
When you constrain the A matrix (that contains 9 regression coefficients) to be equal across two groups, you are estimating 9 coefficients instead of 2*9 in the unconstrained model. Therefore the difference in df should be 9.
In the subsequent moderator analysis you get 2 df because you used the S matrix from the RAM1 object, which constrains the residual covariance between J and T te be equal as well. It will be solved if you use
RAM2$S
instead ofRAM1$S
in group 2. I attached the corrected script.Good luck!
Best,
Suzanne
Thank you so much, Dr. Jak for your response. I appreciate it!
Best Regards,
Srikanth Parameswaran