I tried to plot a DAG(directed acyclic graph) model to find out a sufficient strategy for compounding adjustments(Fig 1）. After plotting the model, I found that data for 'reactivity' and 'inhibition' (both indicate child temperament) are not included in my database, and the pathway from cognitive empathy to prosocial behaviors is biased without the adjustment(Fig.2).
However, if I made adjustments for available variables in my database(Fig.3), the goal pathways(i.e. sleep→prosocial behaviors, sleep→cognitive empathy, and cognitive empathy→prosocial behaviors) turn out to be unbiased(i.e. in green).
This is my first time trying a DAG model. My question is 1)Which group of variables should I choose for adjustment? To choose the former one and state the limitations in the discussion part, or just chose the latter one? 2)How to understand the model and possible biases according to the two different strategies?
Hi Tingyu Rong,
Maybe the attached file can help to understand how topology, intrinsic to the model, can be used to explicate the structure of the corresponding DAG.
The topological-triple is a summary of the dependencies among the vertices of the DAG.
Please let me know if I can help any further.
Oh, I forgot to mention that your model does not need to be acyclic.
As far as I can see, you are trying to put all cause/effect relations among some notions in your graph. Feel free also to have cyclic relations. Topological-triple resolves also topology of cyclic graphs (and its complexity is not NP-Hard). Please see: Presentation Topology and Structure of Directed Graphs - How to Deal with Cycles
Once computed, you can use any one of the three topological orderings of topological-triple to iterate either from cause to effect, or from effect to cause, under the guidance of the other two.
Topological-triple reflects the true topology of the model, and it is bidirectional.
One can also say that it shows the paths of the implementation of "logical implication" operation. Maybe this can give some insight: Presentation Topology and Structure of Graphs - All Paths and "Seven Brid...
Draw a mind map. Feel free to include cyclic implications. Once you get the dependency structure of your model, ask yourself, "What are the effects of these ..." or "What did cause this/these".
Hi Enis，
Thank you for replying! I may need some time to digest your materials :).@Enis Olgac
- Badges
- Science topic
- Similar topics
- Publishing