|
|
Inconsistent Parameter Estimates |
|
Message/Author |
|
toby hopp posted on Wednesday, February 28, 2018 - 1:42 pm
|
|
|
I have a quick question regarding parameter estimates in longitudinal context. Essentially, I'm estimating a series of structural models that involve data measured at two occasions. Model A looks as follows: T1A --> T1B --> T1C Model B looks as follows: T1A --> T1B --> T2C T1A and T1B are the same variables. T1C and T2C are repeated measures. I would expect the unstandardized parameter estimates between T1A and T1B to be identical in both situations; however, they vary slightly (trivially, really) and I'm not sure why. To be clear, I am using data derived from complete cases only, using MLR estimation, and have a couple correlated residuals (identical across models). T1A is an observed variable (treatment/control condition) and T1B is a latent variable. |
|
|
I would not expect those estimates to be the same because the models have different fit to the data. It is not like with all observed variables. |
|
Back to top |
|
|