Attachment | Size |
---|---|
syntax of the univariate moderation.txt [6] | 12.76 KB |
the result of moderation models.jpg [7] | 264.17 KB |
Hi everyone!
I am trying to run a univariate moderation model with binary moderator(smoke or not) and binary outcome(T2DM or not). However, I have doubts about the result. I have run a full moderation model ( I chose the ACE model, because when calculating the heritability of T2DM and smoke separately, the ACE model was the best fitted model for both of them) and several nested models. Nested models are compared with the full model by Chi square test. The best model is chosed by the Chi square test(p>0.05) and the smallest AIC.
Here is my doubt:
(1) In the full model(ACE_XYZ), the moderation on the addictive genetic variance (βa) is not significant( CI contains 0), while in the best fitted model(ACE_XZ), βa is significant. That's so weird.
(2) In the ACE_YZ model, βe is not significant. However, when dropping βe, the ACE_Y model is significantly worse than the full model(p<0.05).
I am wondering what causes the phenomenon described above. Is my result reliable? How should I choose the best fitting model? The result of my moderation model and the synax I used are in the supplementary material.
I'll be really appreciated if anyone can give some oppinions or suggestions!