Is it necessary to test the significance of marker indicator in SEM?
I have a SEM model in which latents have two indicators. I estimated hybrid model using one step approach. When I release constraint one maker indicator (I mean I changed the marker variable) the marker variable has a sign contradictory with theoretical expectations(by the way it is statistically significant)? Is it a problem? If I do not release the marker indicator and do not test if it is significant and in harmony with theoretical expectations, is it ethical?