|
|
Interpreting the intercept in mixture... |
|
Message/Author |
|
fred posted on Tuesday, May 22, 2018 - 9:25 am
|
|
|
Hi, I am running a fairly basic multigroup analysis within the mixture model frame with known class option. Model overall is M by C1-C6; C2 WITH C1; M on Cohab Exp age; DV on M Exp age ; where both Exp and DV are binary variables (0,1) The second group's model has the intercept of C2 freely estimated. C1-C6 are items wit Likert-type response categories from 1 to 5. In the output the intercept of M is repoted as: M estimate= 0.024 SE=0.075 Here is what I am not able to understand. The scale of M is tied to the first item (C1) so it should go from 1 to 5. Also te itercept of M, I am thinking, sould be the value of M at Exp being 0 (and Cohab and centered age at 0). I dont know how to interpret the estimated value of 0.024 (and 0.00 for the second group), especially as I want to interpret the effect of unit change in M for reporting the corresponding OR value/change in DV. So, my question is why is the intercept of Mat around 0 when it should be somewhere between 1 to 5? Thanks for any help! |
|
|
Mean(c1) = a + b * Mean(M), so the intercept a does not need to be the same as Mean(c1). |
|
fred posted on Wednesday, May 23, 2018 - 10:30 pm
|
|
|
Thank you, I see that about the indictor items. My question is more about the intercept of M, since as a mediator in tis case M has intercept and not mean. What is the faulty part in te assumption that the intercept of Mediator at b*0 (Exp at 0) should be in the scale of Mediator's items (1-5). Wouldn't Exp at 0 render the intercept of M equal to M's mean (that is M is not effected by IV and should reflect the mean of factor score)? Thanks again |
|
|
Yes, for exp=0, the mean of the factor m is the intercept of m. But that's not the mean of the factor indicators of m because the indicator equations have intercepts, so it comes back to my formula. |
|
Back to top |
|
|