Degree and Time of Adoption Diagnostic

[aoliveram]

Hi all,

Following up on Tom's suggestion:

“It would be cool to have a diagnostic that returns the correlation between in- and out-degree with time of adoption. 

Could be an indicator of whether opinion leaders were supporters by being early adopters or opposers by being late ones.”

Here's my initial plan:

1. I'll set up a longitudinal panel with binary adoption status over time.
2. Each time point will include degree measures as predictors.
3. A logit model will be used to evaluate how in- and out-degree relate to adoption timing.

I’ve prepared an R script draft that handles panel construction and also provides some visualization of the adoption dynamics.
@gvegayon , what do you think of this approach? Let me know if this fits with your vision.

[gvegayon]

Hey @aoliveram, I think that could be done with some of the functions in the package already. In principle, you could use the as.data.frame method for diffnet objects to extract the data ready for a regression model (see the diffreg function), and then estimate the model. Since time of adoption between individuals is not IID (as they are correlated by construction), perhaps we could add a bootstrap process to estimate the standard errors and whatnot. I'll assing this to the AI to see what it can do. For the AI:

• Ensure to create tests and proper documentation using roxygen2.
• The new model could use the korean family planning data as an example.
• The test should check that the method works. You can do so by reimplementing it within the test and check that you get at least the same coefficient estimates. The bootstrap estimates may be different.