I have used EQS many years ago and when I did the data in each individual variable I believed were independent (they were individual participants' scores on a standardized measure). So I had no a priori grounds to suspect that the value of each participant's data was not independent of each other participant's data.

Is the integrity of an SEM undermined if the opposite occurs? Does it affect the assumptions of SEM if the data in a variable is probably not independent? For example, if the variable was perceived attractiveness, and N = 250 scores were provided by 5 raters (who provided 50 scores each), it seems to me that the 250 pieces of data will not be independent as each of the 5 scorers may have personal biases that operate on the 50 people they assessed (e.g., a rater may prefer tall people, irrespective of any other characteristic).

My suspicion is that this woudl be an issue, but I am seeking to confirm or refute that suspicion.

Thanks in advance for any thoughts..