Message/Author |
|
|
In ESEM and CFA mean testing, the means are compared by fixing the mean of one group to zero and freely estimating the others. Is there a simple command to rather compare the mean of the whole sample to each group. i read some papers using mplus which indicated that they compare the mean of the whole sample to individual group means but just can't figure out how to write such command. |
|
|
There is not a special command for this, but you can create it using Model Constraint. Just express the overall factor mean in terms of model parameter labels and express deviations from that for each group. |
|
|
Thank very much. Can you illustrate for me. |
|
|
See UG ex5.20 for the principle (in a different context). |
|
|
thank. last question for the day. I am comparing means in ESEM with the grouping (36=AUS 76=BRA 124=CAN 246=FIN 392=JPN 788=TUN). why i cannot make 246=FIN the reference group in ESEM. the default is 36=AUS which i dont like. No problem in CFA but ESEM just dont work. |
|
|
Just mention 246=FIN first in your list. |
|
|
Thankx. i was using useobservation command and i realised i should let it be the first as well in the useobservation command other than that it will not work. The subpopulation and useobservation command which one gives the best estimates and the right one to use. |
|
|
SUBPOPULATION is for complex survey data using TYPE=COMPLEX. USEOBSERVATIONS is for all other cases. |
|
|
But for mplus 6.12 The SUBPOPULATION option is not available for multiple group analysis. Or there is an extra command needed. |
|
|
It is not available for multiple group analysis in any version of Mplus. |
|
Back to top |