I'm going to talk a little about, replication, reproducibility. And I think in a lot of, circumstances, people tend to use these words interchangeably. But for the purpose of this talk, [UNKNOWN]. There's going to be a distinction. So, when I think, when I think about replication, I think we talk about like the usual thing that happens in science. The purpose of replication is to address, in my opinion, the kind of validity of a scientific claim, right? So if you say that x is related to y. The first question is well is this claim true? So you have to replicate it in another study or a series of other studies. With new investigators collecting new data using different alloquital methods, laboratory instruments whatever it may be. And you have a bunch of replications and no one can sign, no one can find the same thing now you think okay, well that's probably not true. But if everyone finds the same thing, roughly the same thing, in independent studies, then you think okay, that's probably a reasonable claim. So replication's kind of how we've always done things. And it's still, I think, the ultimate standard for strengthening scientific evidence. So that's replication. Reproducibility in case you haven't heard the term is y referred to as kind of saying the validity of the data analysis right? So here you have situations where data are collected and you analyze with someone else and they say something like x is related to y. So here the idea is that you have two investigators but you look at the same data and, and you used the same methods, perhaps. See if, you know, what they claimed is actually produced from the data that, and the analysis. So the real question, I think the question here is not so much, Is the claim true? But really, Can we trust this analysis? Okay? So that's kind of, in, what I see, the, the key question of reproducibility tries to address. Can we take their approach, apply to their data, and get the same thing that they've gotten? And I think this is key. A reproducibility is, is particularly hard in my area of work in, in which replication is just not possible. So this is where I'm starting. And just so, I think some background everyone's familiar with this is, there's been a lot of talk about reproducibility in recent years. Some of the transars just have, a lot of stages that can't replicated they're huge. I switched edimology and have these 25 year covert studies that just can't be replicated in a reasonable amount of time. There's a lot of technology, which is affecting the rate of data collection. The complexity is very high. You have these, kind of, databases that people are just merging together like crazy, I'm one of those people. And I've seen a lot of these data used off label. So for example you have administrative data sets that are being used for health studies and things like that so this is very common. And then because it's inexpensive computing power is very sophisticated now so we can do very kid of complex analysis on just small data sets. And so, and then for every field x, there is a computational x. So. To sum up the results, so basic analysis can be very difficult to, to describe now, in just kind of like your standard method section in a journal. Oftentimes there's very heavy computational requirements that are thrust on people with very little training. And so just a little bit of the context that kind of, from where coming, I, I think, you think of the range of kind of computational ability from low to high. You know, most of the people in the room are over here. All the people in the room are over here, maybe, you know, Jeff's over here. [LAUGH] But there's a whole slew of people on this end of the spectrum. Who are doing computational, who have to do computational work for, whether they like it or not, and the reasons they've got all this data coming in. So and so that can lead to problems. You can have errors introduced into long-term analysis pipelines. And you can't tell people what you've done in a reasonable amount of space. That complicates the transfer of knowledge. So in many cases, there is a sense that complicated analyses can't be trusted.