How Do You Validate Your Models?

It's not clear to me that there's any real notion of ground truth in some of these measurements. We can't say that we have a high confidence in every single small area when we don't know how much variability or what sort of noise exists around these measurements that we're actually using. The best thing that we can do is be brutally honest in the way that we provide a response. Be honest about situations in which these models don't work as well, or these data might not be informative and really drive towards better understanding this data generation pipe line. I think t it's a multi faceted issue. A lot of people just use their phone and don't necessarily realize the amount of