
Message/Author 

Dan Berry posted on Tuesday, August 04, 2009  6:14 pm



Dear Drs. Muthen, Is there a way to test specific linear hypotheses based on information from the fitted model? This would be something akin to the “TEST” statement in regression or multilevel modeling in the SAS statistical platform (if you happen to be familiar with it). For instance, let’s say I have aconditional quadratic LCM, and the predictor is significantly related to both linear and quadratic growth. I pick out “high” and “low” values for the predictor variable (e.g., mean + 1SD, mean – 1SD) to display the way the curve differs at these values. Within a given conditional fitted curve (say, one for someone with mean  1SD) is there a way to, for instance, test whether the curve between two particular points in time (e.g., time A and time B) differs from zero? Or, for example, test whether the scores at time C of the curve differ between those with high and low levels of the predictor? Etc. Thanks for any thoughts. Best, dan 


You can use MODEL CONSTRAINT to do this. You will obtain standard errors for these new parameters. 

Back to top 

