So a lot of the work that I've done on tuberculosis was using a data set of tuberculosis test results. And so when we think about kind of data quality issues, a lot of them fall into the category of missing data. And so that is missing data in terms of you have individuals that are in your data set. And maybe there isn't information recorded for all of the characteristics that you want to know about them. But also, you can think about missing data as the people who aren't in your data set, the data sets only capturing a subset of the entire population that you might be interested in. And so you want to kind of think about what to do about those two things. And with this tuberculosis data that I worked with, we had both of those issues, and and many more. And so what I was able to do, and part of it was saying, Okay, well maybe we don't have the universe of all the people that have TB, but we have the universe we know all the people who have been tested for TB and saying, well, maybe let's think about the policy recommendations for that group. What can we learn about them? And what policy recommendations can we bring into the game to address that? So in that case, we're actually being creative to abstract away from the missing data problem. And so even if we don't have that data, we're able to reframe the question a little bit, and then develop really good policy recommendations.