I would like to briefly talk about the presentations of research plans last Thursday and Friday.
Your research falls mainly into three classes. "One-channel" empirical models, "multi-channel" empirical models, and "other.
These are models like SERVQUAL where everything goes through a single mediating variable (in SERVQUAL it is "service quality") to cause business outcomes (demand, adoption, continued use).
Based on this image we can imagine several kinds of questions that the AG might ask.
Most important, "why do you think this model represents important relationships in the real world?"
Why would somebody care about "satisfaction"?
Can you measure "quality" independent of "satisfaction"? How?
There may be a direct measurement of "quality", such as "waiting time" for a call center.
Without a direct measurement, an instrument such as "price" is needed. (Usually as price goes up, satisfaction will fall.) If you have no measurement or instrument for the mediating variable, the model is "not identified".
Can you control some or all of the "factors"? If not, the model may be interesting to you but it is not of practical use. (In applications like marketing, "control" includes "select on", as in market segmentation.)
The model itself may not be persuasive to the AG. Here is a different model:
In this model the AG objects that (1) not all "factors" cause "quality", in fact "quality" causes "not a factor", and (2) there are "new factors" that you didn't account for.
You need to be able to justify causal direction, and deal with such proposals of new variables. Best is if you (or a previous research) thought of the new variable, and you can give an answer why not to use it as a factor.
"Taking the model too seriously" means thinking about the effects in your model as if
This is the main cause of "having no answers."