I found myself surprised by the fact that some epidemiology is based on "impressions that people have." Megan Murray noted that most of the field is data-based, but sometimes there are still studies based on hunches. I thought that epidemiology was a strict science without room for hunches at this point. However, I guess it makes sense that some of it would have to start with first-hand observations and hunches. I just find it surprising that we haven't somehow moved past this, despite it being a perfectly reasonable way to form a hypothesis.
If I had conducted the interview, I would have asked more about what an "impression" looks like. I'm really intrigued by what counts as a scientific impression versus just a coincidence that you happen to notice. I wasn't sure if the example of a salmonella outbreak would be an example of that, but it seems like it could be. I'd also like to know how many epidemiological studies start with an impression as opposed to a data-driven hypothesis.
I'm curious as well about your question about "impressions" -- it led me to ask myself about the relationship and differences between causation and correlation. With the salmonella example, how do epidemiologists assert that the spoiled food is the root cause, where they are not able to get hard evidence about the contents of the spoiled food after the fact? It seems obvious, but this is an extremely clear case, and I'm curious about where the line is drawn when situations are more nuanced.