 OK, so the idea for this this conference really came from a lot of discussions we had in each of the different ES hackathons, particularly where we started off in Stockholm. Thinking about how we can get evidence that this workflow is better embedded into INTA. So the first step really was to think about what is there, what's available. This is actually quite a tough thing to be able to do, but luckily for meta analysis at least we can find the CRAN task view and what that does is it points to packages that are being described by the maintenance as meta analysis packages and there's about 150 packages on that CRAN task view. You can see that the number of packages has increased over time from Pioneer here which was meta, meta, so very general package between meta analysis all the way through to the latest package on the list is NMA plateplot that's from very general but also quite specific packages as well that have been developed to help with meta analysis. Obviously the number of downloads, there's lots of biases in these but I wanted to find a way to really look at the other packages that are sort of rising above from what they would expect by the relationship between the number of the ASIMs and the number of downloads. Some of them are, some are quite astounding outliers. So up there in the top left hand corner you've got FXIs, that's a package that has probably the widest range of FXIs calculations I've seen, both Bayesian and Frequentist. Inter remains, it's quite an important one, it's used by, it allows the data format from Revman, so it's obviously quite important for systematic reviews. MetaBMA, another fairly relatively new one but also highly downloaded, looks at Bayesian model averaging for meta analysis, so really some of these things I'm going to be saying in this talk, it's about what is it that for you guys, you think, leads to a package being highly downloaded, how much you use it, so how are they all related? So for this bit I need to just come out of here, and come up here, it's an option mode and then we can have a look here. So I took the CRAN task view and I've looked at the dependencies of the different packages that they do there, so the blue ones are the meta analysis packages and the yellow ones or the gold ones are what I've called sporting packages, so they're packages that are like GGplot for example, which underpins some of the functions that lie within the meta analysis packages. There's about 500 of those sporting packages and you can see that they're by their sizes, the number of links that go to other packages. From the meta analysis packages side, not really surprising but basically all roads lead to Metafor. So a lot of the functions that underpin other packages in meta analysis, meta analysis packages are actually come from these Metafor packages. Metafor itself is linked to MLME. So Metafor is by far the largest contributor to other packages and you can see that even where a package doesn't have a direct connection, it's often connected through other packages as well. So Metafor has really facilitated a lot of the functions that are happening in these new ML packages. Let's go back to that point, let's read that so I can see what I'm saying, okay. So here we've looked at how they're related, so what do these packages do? So as a first point, let's start, we started to look at their package descriptions. What is it they say they do? Most of these packages, so this is a plot of the words in the descriptions and how they're related to each other and the size of the arrow is related to the number of times that information. So a lot of these packages about 40% are dedicated to running Metafor analysis. But these are for sort of different data types, so for survival data, for genetic data, correlations, you name it, there's a different Metafor analysis package for it. 11% of these packages are to do with effect sizes, converting effect sizes, yeah, I'll show you my pointer. And so there's quite a few of those. 7% of packages, sorry, are to do the sensitivity analysis here. And 7% of plotting functions mostly to do the final plots, but so are a few other 10 plots. And 6% to network meta-analysis, so there's quite a few of those that are coming through quite early in the Metafor analysis here. So what's next? These are sort of things really that are more like discussion points and perhaps questions for you, rather than the things that I should be doing. But one thing I'd like to do is update the current tasks for you, but adding these non-crime packages too, but I need to find a way to actually systematically find these specific packages. I would like to provide guidance to users on which package to go for, so if you have a specialist interest in a certain area of meta-analysis or you have a generalist interest in meta-analysis, where should you go? Which package should you go for? Over the next couple of days, I'd really like to start thinking, and for you to start thinking about identifying the gaps in the function. So what are you missing? What's not in these functions? Do we need your packages or can we just sort of incorporate those missing functions into assisting ones? Related to that, we need to identify a functional redundancy. So what do we have that we've got a lot of and don't need any more? I think FXI's conversion function is probably one of those things. There's lots that are really good in other packages, so we don't need to reinvent them. And finally, I'd like you to think about what makes a meta-analysis or ECOSPEN evidence-success package usable? Is it the documentation? Is it how versatile it is, how general it is, how specific it is? What makes a package useful for you? So if you have some opinion on that, you'd like to stick some information in the Slack for me, or you can send me an email, whatever you like, how are you on today. But I'm really just interested in your views on those things. So thank you very much and enjoy the conference.