My first question relates to how to measure the volume of conversations and classify the types of conversations. Maybe I miss the information in this paper, do you assume that all the relevant conversations occur in the Tow Square and they are incidental? If that is true, could you explain the rationale behind the assumption and how exactly the volume of conversations can be observed as I think one conversation can be fragmentary in nature and doesn't have to be coherent and continuous.
My second question relates to how you develop your current model given the fact that some previous network communication models don't apply to your research design and assumptions. In your current model, the theoretical implications based on assumptions and proofs are consistent with empirical results. I am very interested in the back and forth considerations you take in the process of developing your model between research design and empirical analysis. For instance, when and how you make assumptions about the prior distribution F(v) in the model.
shugamoe
commented
6 years ago 
Thank you for the presentation.
My first question relates to how to measure the volume of conversations and classify the types of conversations. Maybe I miss the information in this paper, do you assume that all the relevant conversations occur in the Tow Square and they are incidental? If that is true, could you explain the rationale behind the assumption and how exactly the volume of conversations can be observed as I think one conversation can be fragmentary in nature and doesn't have to be coherent and continuous.
My second question relates to how you develop your current model given the fact that some previous network communication models don't apply to your research design and assumptions. In your current model, the theoretical implications based on assumptions and proofs are consistent with empirical results. I am very interested in the back and forth considerations you take in the process of developing your model between research design and empirical analysis. For instance, when and how you make assumptions about the prior distribution F(v) in the model.
Thank you so much.