A collection of tips to ensure reproducibility for R-based clinical submissions, a modern architecture lands in R-Universe, and going beyond the built-in mtcars set to educate students with NYC Open Data. Episode Links This week's curator: Sam Parmar - @parmsam@fosstodon.org (Mastodon) & @parmsam_ (X/Twitter) Making R Submissions Reviewable for FDA Windows ARM64 comes to R-universe The Journey of {nycOpenData}: From Classroom to Community Entire issue available at rweekly.org/2026-W33 Supplement Resources R Consortium Submissions Working Group https://rconsortium.github.io/submissions-wg SimpleEnglish Agent skill https://github.com/AminBlg/SimpleEnglish New R tooling https://r-devs.com/ Supporting the show Use the contact page at https://serve.podhome.fm/custompage/r-weekly-highlights/contact to send us your feedback R-Weekly Highlights on the Podcastindex.org - You can send a boost into the show directly in the Podcast Index. First, top-up with Alby, and then head over to the R-Weekly Highlights podcast entry on the index. A new way to think about value: https://value4value.info Get in touch with us on social media Eric Nantz: @rpodcast@podcastindex.social (Mastodon), @rpodcast.bsky.social (BlueSky) and @theRcast (X/Twitter) Mike Thomas: @mike_thomas@fosstodon.org (Mastodon), @mike-thomas.bsky.social (BlueSky), and @mike_ketchbrook (X/Twitter) Music credits powered by OCRemix Chillout - Mega Man 2 - sedure - https://ocremix.org/remix/OCR01175 Brushwork - Mario Paint - Fratto, Wiesty, The OC Jazz Collective - https://ocremix.org/remix/OCR04389
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Hello friends. We are back of episode two thirty eight of the Our Weekly Highlights Podcast. This is the weekly podcast where we talk about the amazing highlights and other great resources that are shared in this week's issue of R Weekly, which you can find at rweekly.org. My name is Eric Nance, and I'm delighted that you join us from wherever you are around the world. The month of August has a lot of moving parts, as the cliche goes for what I'm up to these days, but luckily this is my sanity, thirty minutes or so, because I get to be joined by my co host who knows a lot about what I go through and has his own share of stories to tell. Mike Thomas. Mike, how are you doing today?
I'm good, Eric. We're all surviving, not thriving or thriving and surviving depending on the week. You know how it goes, probably more the former than the latter. But we got some great highlights today that I'm super excited about. We sure do. And, no outside platforms will do, wreckage to us like a certain one that starts with G did to me earlier this week and ends in B, you fill in the blanks, folks. But we don't have to worry about that here because we got a great issue of our weekly to talk about and this week's curator or this issue's curator for this week is Sam Palmer,
another close friend of mine especially in the life sciences sector, and as always he had tremendous help from our fellow RWQY team members and contributors like all of you around the world with your poll requests and other suggestions. I just mentioned there are a lot of moving parts this month, especially for me, but I could say for probably most of this year I have to put on my other hat, so to speak, for a second here because on top of doing this humble little show in my day job as a statistician of life sciences,
I am also, one of the longtime members of a working group that is trying to make the R language easier to use in the realm of life sciences and specifically submitting clinical trial results to health authorities, most notably at this time the FDA here in The U. S. We've made a lot of progress over the years. We started with pilots that just did very simple tables in R. I did shortly thereafter, did a Shiny app in R,
transmitted as an R package with, you can guess, the Golem framework, because, hey, if I'm putting a Shiny app through the gateway to the FDA, I'm going big and that was a lot of fun. We've also had transfers of data sets created in R, these intermediate formats, one of which is called ADAM that gets used a lot in submissions. We've had some pilots also about the use of novel technologies such as containers and web assembly,
and more recently looking at other formats for the kind of data that goes into a transfer that's meant to be portable of sorts, I. E. Looking at the JSON format. So all these pilots have taught us a lot. With that said, there have been challenges that have bubbled up from time to time, and we definitely have seen some challenges this year that is surfacing because now we are seeing in life sciences R is being used more in clinical submissions. And with that, there's a bit of variation involved in how
these submissions are being sent to our colleagues at FDA. And that leads me to our first highlight for that long winded introduction. We have a blog post from the R Consortium, which is the umbrella organization that our working group is under. And this is about tips that we have collected over the years of our pilot experience, as well as very direct feedback from our FDA reviewers on how we can make our submissions more easily reviewable
by health authorities such as FDA. This gets to a few different areas, and I will promise you I will keep this pretty brief because if you give me an open mic, I would probably talk about this for five hours because I have so many thoughts. But I'm going boil down to a few key points that this blog post highlights. There are issues that can happen with technically getting the R environment set up and to be able to actually run the R code,
irrespective of what that code is actually doing. Just getting to that starting point. There's a lot of variation involved. But there are also things where reviewers have more experience potentially with a certain language that has three letters, and they just might not have as much experience with R and want to know why certain things can be different. And with that, it brings in some recommendations that this post, you know, discusses here where you, a sponsor, I. E. A life science company,
you can make it easier for the FDA reviewers, but also I'll comment some existing challenges that we're navigating through. This advice, whether you're in life sciences or not, I think is still going be pretty appropriate, especially if you're in the realm of reproducibility and collaboration to adhere to reproducibility principles. One of these is communicate early, and there is a formal cadence to communicate when you submit a trial to the FDA before
you get to that point where you've got the data and output ready to go. That's where it is time in those early meetings to say, if you're a sponsor looking to use Hey, FDA reviewers, we are going to use R for this. We want you to be aware of it so that we can have a dialogue of what you are expecting and things that we should watch out for. That is a non technical thing. That is just good advice across the board. Now we get to some of the technical stuff. There is one of these that I feel,
if I have to put not only my R Consortium working group hat on, but also my I am a Linux geek of all things technology hat on, it should be very easy for the FDA reviewers to install R and reproduce that environment. There are certain things that we know as of now what happens. For better or worse, they are looking at R on their laptops, are looking at submission on their laptop that FDA provides to them, and it's up to them to put R on it.
Therefore, we should have very clear instructions and test those instructions out beforehand before that goes through a gateway into the FDA reviewer's hands. There has been a lot of variation around that. I'm going to zip my lip there because I could say more, but I want to keep my job. Nonetheless, let's get to the next ones here. There can be situations where the code that's being sent has been developed on a different operating system, at least in where I work and many other bigger companies,
there is a shared Linux environment, maybe it has POSIT Workbench on or some other R infrastructure on there, and that's where their code is being developed. That may be where the data is as well. Well, when an FDA reviewer is going to get this submission package, they're not accessing that same system at the sponsor. They're using their own laptops to do it. So you got to watch out for potentially operating system differences and of course just where data is located. You got to make sure that the paths to said data files
are portable enough or that you've tested that before you send it through the gateway because they're not going to have access to that huge file system perhaps that your data resides in. And last one that is more recent, but yet makes sense when you hear it. If you are reviewing a result, just put yourself in the shoes of an FDA review for a second, and you may have a question about a certain result. Not saying it's wrong, but you may have a question about it. You go to the code that is associated
with that table or listing that shows that result. You just want to get to that one maybe linear regression model p value or that t test p value or comparison, pairwise comparison. You have to, typically speaking, muddy or wade your way through maybe a very complex R script of functions nested in functions, nested in functions, often with probably over 20 parameters
of which maybe only one or two are relevant to you, and you just want that one result. You don't care about how the table is generated. You don't care about the listing, how that listing is generated. You just want that result. So one thing that would be very helpful in these cases is on top of that code to get to that actual output that they're viewing, you have maybe an accompanying script that just gets to those statistical results, and it's easy for them to find, easy for them to debug.
There are efforts to make that easier in terms of leveraging some summary type of data format, one of which is called the analysis result dataset or ARD, that could be a big player here, but that hasn't had mainstream adoption yet across the board, but I think it is on the way. So with this, at the end of the post, there's a nice little checklist that if you are working at one of these companies and you want to take this advice, it may be something you want to take to your respective study teams and say,
Hey, if we're using R for this submission, maybe we need to look into this. So I think there's a lot of things happening in this world I'm in with trying to make R more readily submittable to the FDA. This is one of the byproducts of the fact that we've made good progress, but there are still some fundamental issues that if we work together, I. E. Life science companies and regulators, we can make it better for everybody,
accounting for certain challenges that are not solved just yet. But I've always been of the mindset, no matter how I'm talking about this domain, I'm talking about the stuff I work on on a daily basis, you control what you can control, and there are certain things that we can most definitely control and make better. So there's a lot more in this space to come. I will be talking a lot about this in upcoming conferences, as well as some future pilots potentially to address some of these concerns, but it
is a reality check of sorts. We've gone a long way, We have more to do, but we will not be able to have this dialogue if not for this working group that I'm a part of. It has been very helpful, and we don't want to stop this anytime soon. So yes, very life sciences specific, but I think there are some nuggets that if you find yourself in a similar situation with working with other teams or a high profile analysis that is, you know, highly important, has a lot of rigor behind it, a lot of these principles I think you can take to your respective organization.
No, I couldn't agree more. And, know, this blog post is veiled under the guise of, you know, pharma and FDA submissions. But the whole thing to me, as somebody from the outside that's not in the life sciences space is just good software engineering best practices, right? Which are sometimes hard hard to come by but important in my opinion in every industry. Not just not just life sciences. So, I think there's some some great takeaways here.
Obviously, writing code that's written for reviewers is super important. I think one thing that can sometimes help obviously is Roxygen style documentation. Maybe that's helpful, maybe not, depending on sort of who your audience is and what they are used to. But also, I think nowadays, in terms of how your code itself looks and is formatted, we have this great new tool called Air that you can just literally click the button. And it will consistently
format your code across the project in terms of spacing and and breaks and number of characters and things like that to make everything look, you know, very co cohesive in your whole project, which I can think which I believe goes a long way towards, you know, other people taking a look at that code and understanding it quicker. And obviously, something like a single obvious entry point for reproducing key analyses, that's one that we've Eric, you and I probably harped on for years now at this point along with many others. You you shouldn't have to run script oh one and script oh two then script oh three. You should have some sort of a a main. R script that that governs everything or I think maybe the smoking gun
or maybe smoking gun is not the the the elephant in the room here is is targets probably applies in a lot of these cases. And can be really helpful especially for folks who are trying to do some of those things that you talked about, Eric, like, trace out one p value in a computation, right, in in some sort of a result and figure out, you know, what are the dependencies upstream from that that I need to care about and which are not. And the targets package can give you literally a dynamic visual to be able to trace that out that,
allows you to really quickly, follow the logic that got you there. So I think that that's fantastic. This brings me back to a book, I think, that Bruno Rodriguez wrote probably a few years ago now. Right? Building reproducible data science pipelines. I don't know if that's the exact title of that book but I think I'm pretty close and that has all of the same sort of advice you know, sort of fantastically wrapped up in a very digestible book for you that I keep very close to my desk on a day to day basis.
One last thing that I'll I'll point out sort of in the closing of this blog post, talking about the increasing adoption of like, LLM assisted programming and making it easier to produce code that's sort of overcomplicated, deeply nested. One thing that that may be helpful for folks out there, especially in these sort of technical spaces that we've adopted in our team recently, is this Claude skill. I would hope that you could adopt it even if you're using, you know, OpenAI Codex.
I believe it it works in sort of harness there, but it's called simple English, and we can put a Oh, I'll I'll definitely link to that. Very good. Yeah. We can put a link to it in the blog post. I think it's, like, the number two repository on GitHub lately or something like that. It's gotten very popular, but it follows this this English convention for writing manuals, I guess, like writing airplane manuals. And out of folks that wrote airplane manuals, they developed sort of this whole methodology called, I believe, STE
certification. And there's some rules around it, only saying the thing that you want to say once in the sentence, not necessarily repeating it. And it has helped us not only reduce like tokens in with these models, but just make the pros and the narrative that we're writing or the in the code comments that we're creating just much more straightforward than what you get with Cloud Code and others that seem to be overly verbose. So we'll link to that. That may be helpful here.
But awesome awesome blog post. I'm always a sucker for any blog post that's talking about reproducible data science because that is near and dear to my heart. And in this industry I'm in, in particular, it is so important for many levels. My hope, and this is where I get to my future ambitions here, we've got a couple of key areas that I think we're going to start making real headway in. A, trying to get to the bottom of this environment bootstrapping issue. I mean, I wish containers are more feasible.
They're so close, but there's one very big limitation we're dealing with on the FDA side that we haven't found a good solution for, but there could be ways in between to make this more seamless. But it's also making sure that if there are questions about how R does things, especially on the statistics side, that we can take a closer look at that and provide reassurance that, you know what, this has been battle tested, even if we have to make our own simulated data sets to prove it. You know, I'm not opposed to doing that. Looking at maybe another language side by side with it, so be it. I think we can make some significant progress there. So more work to do, but again, I'm very happy that all this is being shared
transparently because until we had that first Our Pharma Conference back in 2018, all of us in life sciences were kind of talking this within our walls. Nobody really knew how to talk to each other about this, but now we're here. So hopefully, hopefully this is the start of some even bigger transformations down the road. And yes, yours truly probably will have yet more hot takes about that as we go.
And up next in our highlights, well, it was a very high praise of you, Mike, in that first segment to talk about what the virtues of say the targets package could be in the situation of reproducible code and tracing back to different steps in analysis pipeline. Well, where does the targets package sit? Well, it sits under the ROPENSci umbrella and, of course, with targets being offered by Will Landau, but, of course, ROpenSci has many extremely high quality packages in the realm of science
and data science in general, but that's not all they're known for. They also have very cutting edge efforts in the world of infrastructure to help package authors really test out their packages across so many different permutations of, say, R versions, platforms, and more. And one of these efforts that has brought all this together is an infrastructure paradigm that we talked about for over a year, if not longer,
on this show, and that is called R Universe. If you don't know R Universe is, you'll definitely want to check it out because that is where you, as a package author, can simply add one little thing to your GitHub profile, and then every commit you make or push to a branch that you make for your package, you can have binary versions created of your package under the R universe and have an alternate way for users to install your package. It is really awesome.
And as I said, it supports a lot under the hood here, but there has been one other newer ish platform that just as of last week now has some really nice support and that is the ARM64 architecture under Windows. So this blog post comes to us from the architect of RUniverse, your own Ooms, announcing that Windows ARM 64 has come to RUniverse proper. You may be thinking, wait, I thought ARM was already there. Well, was for macOS and Linux, but Windows has been a little bit later to the game, but it is here as of now.
And a lot of times, in my early days of using packages in Windows and even developing a package in Windows, we would only have one thing to really rely on, and that was what the CRAN servers had in their infrastructure, wherever that happens to be, of testing Windows builds and hoping for the best that everything worked. Well, R Open Sci has been helping with Windows in general for a while with R Universe, but now with the ARM platform,
they bring this to you for support, but it wasn't all easy to get there. There were certain things that had to happen in order for R Universe to adopt, the ARM64 architecture for Windows, one of which is making sure that with the R installers that you use to install R, that it can pick up the ARM64 binaries from the right repository path when you're using R.
Also, were some changes needed to external software, one of which a Bayesian, a well known Bayesian program called JAGs, needed some tweaks to make sure it could compile for ARM64, Windows. And then also some of the newer packages that are leveraging Rust to get performance in their R package. There was some patching I had to do to make sure it was targeting the LLVM that was combining with Windows ARM and not a hard coded path,
and some other packages that required a lot more compilation need a little bit more massaging, if you will, with a compilation flag. But everything is working good and there's been testing that's being done, but Yaron mentions that you do need to, they need some help to make sure there aren't any gaps in some of these packages being compiled. So if you are one of those users using a newer Windows machine with ARM hardware, your own would really like you to take a look at some of the packages on R Universe
and being able to see if there's any gaps that you can give feedback to your Roam and the rest of the R Universe project and have those solved. But this now goes to show you that if you're using R Universe, you're covered on so many platforms now that should give you reassurance that your package is gonna work just about everywhere except perhaps Solaris. That's a joke for the old timers out there. Is that still in the crayon checks? I think it still is.
Oh my goodness. Unbelievable. Yes. I think you gotta think how far we have come now in Arm 64. There's actually it's a funny story. I mistakenly for a new hire on our team about a year ago, picked him up a Windows laptop with Arm 64 architecture on it when the rest of us were running AMD. So we've had to kind of navigate some of these little gaps in some of the software that we use across the data science ecosystem that plays a little bit differently between the two architectures.
And it's been interesting to navigate some of those differences and and work through them. It's it's been more interesting than it's been frustrating which is good but it's exciting to hear that arm support is coming to our universe because we are big our universe champions. It's a phenomenal project that I think has been pretty transformative for the R community over the last few years. So, super exciting. I think one of the things that probably excites us most here is
I one of the big packages that we tend to use on on a lot of projects where we're using significant whatever you wanna call big data these days but that would be the the arrow package and I know that that was one that the R Universe team had to do, you know, one of these sort of workarounds for because it used compiler flags, that were incompatible initially with ARM chips. And and I have to imagine that that was a pretty significant undertaking for the R Universe team. And I wonder how this all flows into, like, the downstream
dependencies of Arrow. I'm sure there's quite a few R packages, right, that install Arrow as, one of the the dependencies that it uses. So, very impressive that they're able to do this. And I have to imagine, I think we have to say it, right, that probably AI assisted coding helps solve some of these things a lot quicker than we may have been able to previously. But I don't know if that's just a guess. The folks at R Universe are probably
way better software developers and coders than I am. So maybe this is just sort of bread bread and butter for them. But, really, you know, exciting update to the R Universe. I think this is a pretty big deal, as more and more Arm architecture hardware is distributed across the data science community. Would imagine in the next five, ten years or so. I was looking at the stock price for Arm year to date. It is more than doubled.
So I would have to imagine that as that company continues to succeed, then we will probably see more ARM architecture hardware on the market. I mean, I think, you know, you can also trace back to when Apple introduced the ARM architecture for their recent MacBooks, the M chips, M1, etc. That kind of started the conversation more for mainstream and not just ARM for hobbyists and their Raspberry Pis, right? Now it is becoming much more common
in their sectors, but, you know, now Windows has it too. So now all three major operating systems are dealing with ARM in one way, shape, or form. And another important thing about this is that we have no idea when the R Core team is going to adopt Windows and ARM64. It is not there right now, which is one of the reasons I mentioned in the outset. There was some patching required to how R is installed on these build systems that R OpenSide is doing because
to get these packages going, think there had to be some work to make sure it could talk to those binaries. So I've never used Windows ARM, so I have no idea how difficult it is to get started with it, but I think this could be a positive trend to other infrastructure perhaps adopting this in the future. But yeah, I think if you're a package author, it's almost like, maybe a hot take, another one here, you almost have no excuse not to use R Universe. It is just so easy to get on board and it's going to solve so many problems for you. Both your company, Mike, and my open source packages
have benefited so much from it. It just, the time is now to give it a shot. Couldn't agree more. And I know you said Raspberry Pi, but is Mac mini the new Raspberry Pi? Oh, yeah. And it ain't cheap either. No. A lot more expensive. They're a lot more expensive, but, yeah, they are going like hotcakes now. So, yeah, that that is something that is not it's not gonna be on my, holiday, wish list. But but No. I'm not gonna open claw my whole life, but, for the folks that want to, go for it.
And rounding out our highlights, we're gonna get back to learning mode in the classroom, so to speak, in terms of a novel way to make teaching data science a lot more relatable to your particular domain at hand and a fun story around this. So this last highlight actually comes from, just what? The R Open Sight Blog. Again, they're everywhere on this show, and it's titled The Journey of NYC Open Data from Classroom to Community.
This has been authored by Christian Martinez. He is an adjunct professor at the, Brooklyn College over in New York and also the founder of the NYC Open Data Initiative. So let's set the stage on how this all came to be. Well, at the aforementioned college that Christian is teaching at, they have, you know, a program in psychology research and they have a nice class that if I was in that program I would definitely want to take, apropos to this episode,
reproducible psychological research, and they're learning how to code. Yes, I love it. And it's a virtual event or a virtual classroom, it sounds like. So you always, whenever you do a virtual workshop, conference, or an online course, you never really know how much they're engaged. I hope they are, but sometimes you have to hope for the best. Well, in early versions of this program, they were using what KAM and BASE R, the
often used, in my opinion, overused MT cars data set. Yep, it happens. Well, yourself in the position of one of these grad students. They're in this, you know, psychology, psychological research, you know, domain. Do you think they really want to use a data set about automobiles? Maybe some are enthusiasts, who knows, but they probably want to see some that's more applicable to whether it's their domain or at least where they are, something that seems more tangible to them.
Well, that's where Christian here realized that, yeah, it was just not working well for his students. So how to get them more engaged, how to get them more interested. What is a common denominator of it? Well, for this, all of them or most of them are in New York, New York City to be exact. So if you're in a particular area in the world, wouldn't it be nice though, use data from that particular area in the world?
And that is where the NYC Open Data Portal comes in because it's been for years now, apparently, making publicly available data available in this portal with not just hundreds but thousands of data sets. It goes from all the way from transportation data to even crime, but or calls to help services, much, much more, but it's all free. No gate in front of it, and this seems like a really good fit.
Now, at the time that he was exploring this in terms of using it for the class, there were only two methods to bring this data into, say, a course that's using R at the back end. A, export that data as Excel or CSV files, okay ish, but sometimes those exports are outdated. Sometimes they are updated in the back end, but maybe the exports of it have not been updated yet. And also it just is more overhead for the students to put them in the right place, the file's in the right place, make sure their paths are working right. If you're new to R, this can be a bit difficult to manage,
but also they really wanted to get to the most up to date data. What was a way to unlock that? Well, how about leveraging APIs that you could somehow get to the actual upstream version of these data without having to download anything? What does this sound like to you? This sounds like a package in the making, and that's exactly what happened is that Christian saw him, you know what,
it's time to build a package to do this. So he took the matter in his own hands and built it himself. And the first version of this, he thought the easiest way to do this was to just simply have one function to get one specific type of data, three eleven data to be exact, this call data. So he called this function NYC underscore three eleven, which was simply go to that online, you know, portal, grab that particular data, and read it as a data frame. So seem all good.
Well, why he thought, well, if I want to get this on CRAN, is really a package of just one function going to cut it? That's when he thought, you know what, how about I just simply duplicate that function a whole bunch of times, 30 to be exact, for the different types of data that were most popular. Does that sound like a lot of copy paste? It was, and it was a little difficult as someone in Christian's shoes to do this on their own, but it was working,
but they in this pack, these projects associated with this data, they were, the students were able to use this package and they could use it to answer a simple question of like, you know, defining your own question of interest for what they want to do with this data. And then they even put like an online book out there, this open data snooze and gallery book where each project kind of is a showcase of their work. And all this was worth it at that time,
but as this is evolving, he had another idea. What if it's not just to learn about R with this package? What if they actually want to learn about the data and actually contribute back? Maybe some insights about this that could lead to some real help. And what if one of these students wanted to help the package itself? How to make it easier to contribute? And as he's assessing this, he's assessing these contributions,
he decided that in order to get this in the best quality possible, put it through the R Open Sigh process. And that is where in R Open Sigh to get a package on there, it's a very, very transparent, rigorous peer review process where you have an assigned reviewer and they work with you as a package author to work through various issues that may not be errors per se, but there's maybe ways to do things better. And with that, there is a lot of great advice. And one of those,
there was an easy win, so to speak, to turn those 30 functions that were mostly copy paste a code of pulling data from slightly different sources, turning those like 39 ish functions over to just three so that over 2,000 of these data sets could be pulled in this package. That is amazing. That is, again, for this reviewing process, in fact, there were a couple of reviewers here,
Hal Lin Dong and Michael Pascale, that were assigned to this. And through that, after those reviews are complete, NYC Open Data, the package, is now under ROPensci. So in fact, it is dovetailed into additional packages, being open as well, such as New York State open data, looks like Los Angeles open data, Chicago and Austin. So all of this started for education, but no matter if you're in a university course or you are just learning R on your own, but want something a little different than a data set about flowers from the 1950s
or data about automobiles, I think this can go a long way to being more relatable and making that learning journey of R a lot easier for yourself as a learner, and if you're a teacher or a practitioner, making it easier to teach R and getting them really interested in the process. So lots of great story about starting from a small idea, taking it further, taking it further, and now we got ourselves a great quality package to access this open data. So good stuff on rOpenSci blog as usual and congratulations to Christian
for sharing this story and the momentum he's produced now with making NYC Open Data a full reality to all of us. Eric, sometimes this podcast comes full circle in the most beautiful ways, and we started out with software engineering best practices for the FDA, and we're ending with our open side helping out open source R package development just be a little bit cleaner. I'm sure a little more organized, a little more modular into all of those sort of nice practices that we hope to have when we're developing our packages that can go a long way. And I think one thing that you find sometimes when you make those improvements in sort of modularizing your package
and bringing in some of those software best practices is that suddenly, the logic that you've built is maybe even more extensible than you might have thought in the first place. And I think this is exactly what happened in this case as, you know, since that original package was developed, the NYC Open Data one, there have been four additional packages built turning one package essentially into an ecosystem now for Metropolitan Transportation Authority, Open Data, the MTA, Los Angeles, Chicago,
and Austin, Texas, I'm assuming, Open Data, which is really, really cool that you're able to take kind of this one project, put some best practices around it, and quickly find that it's extens ible in a lot of other cases. I think that's something that I've seen play out many, many times. And obviously, there's a lot of folks involved here from the R OpenSide team, from Christian himself sort of taking the initiative. And he has a big thank you to his students
at the end of this blog post as well, saying that this work wouldn't exist without them and their willingness to try something new, contribute code, ask questions, and occasionally tolerate his crazy ideas. And I think that, you know, especially if you are someone in the sciences, studying computer programming at all, being curious is the most important character trait, character trait that you could have and it's easy to get impostor syndrome but if you can develop that muscle to ask questions that might be a dumb question and just not care,
you will learn much faster than than any of your peers and you will you'll stand out in this industry. I can I can guarantee it? It took me a while to get to that point where I am sort of unafraid to to ask any questions but that's really the only way for me to learn to connect the dots and to be able to progress in my own learning and I guess this is sort of a personal tangent at this point but I think it matches the spirit of what this blog post is sort of wrapping in here at the end. And that's my two cents of encouragement for the students out there who might be navigating this industry in kind of a tricky time.
It very much is a tricky time for students across the board here. I fully, fully agree. What we can do as a collective of those of us that have been learning a language like R and been through data science or statistics courses to make it more human, make it more approachable, teaching them the skills that they're, I hate to say, no large language model is going to teach a lot of the stuff that these students are learning with a hands on experience
of not only consuming what Christian built here, but then contributing back upstream to either that package or launching these new packages. Those are skills. Whatever these students use R later or not, those are things that can be transferred across the board. So I have great admiration for people like Christian. Also another professor we feature a lot in our weekly, Andrew Heiss, of how progressive he's been with his teaching and his methodologies. There's so much that is available with these online resources.
It just takes a bit of creative thinking to think about just how this can benefit your students, but also the community in general. I think that's the kind of learning that if yours truly at this advanced age was going back to school, that's the only type of learning I want to be a part of. Absolutely. And if you want to learn more, boy oh boy, every issue of our review has so much to learn. As always, when I look at the issue after it's been released, I see a few things I'm like, okay, somebody's doing that. I want to learn about that. There is all sorts of these ideas to choose from. In fact, one of the things we don't have time to dive into today, there are some really interesting updates
coming to the Shiny packages in R and Python, and some of them are AI related that seem to be a trend that our favorite little package called Golem has started. So you'll learn more about that if you read that glimpse newsletter from POSIT amongst many of our package updates, tutorials. We have, lots, lots more to choose from whether you're in academia or industry or anywhere in between. So we always value what you are seeing in the community. If you're reading a new package or you found a great new analysis that you want to highlight,
we're just a poll request away to get that into our weekly. And how can you do that? Well, if you're on GitHub and it actually loads for you, which hopefully it does, we're just a poll request away. Just do a markdown of that highlight link or that, link to that resource. Our curator link will get that into the issue. Again, assuming the platform's up, you can tell I'm jaded from that. Cheap shot. I couldn't, I couldn't resist buddy. It's been a, it's been a spicy Friday, but what's also hopefully not too is where you can find me, hopefully not causing too much trouble.
Where is that or how to get in touch with us on the show? Well, one of which is very easy, the contact page in the episode show notes. It's right there. Just send us a quick note. I will be sure to get that and run respond back to you or even read it on the show. Also, you can find us on the social medias out there. I am on blue sky where I am at our podcast at bsky.social. I'm also on Mastodon or I'm rpodcastswellrpodcastsindex.social. Too many handles buddy, too many handles. Plus I'm also on LinkedIn. Search my name, you'll find me there. That's the easiest way to summarize that one And also a special,
a special shout out, from our, the author of one of the highlights we talked about last week, Matan. He was very appreciative of our review of his blog posts. And again, sometimes I may say something that may be off the cuff but who knows it might inspire a blog post in the future but, regardless Mike where can the listeners find you absolutely can find me on blue sky at mike thomas dot b s k y dot social or on LinkedIn if you search catch brook analytics, k e t c h b r o o k, you can see what I'm up to.
And speaking of feedback, I literally just saw this, appear. I didn't see it till just now. I want to thank Noveka that sent a message on, who sent me a message on Blue Sky, an additional resource to see all the various tools around R, and this may be related to, we were talking about this earlier, Mike, tools like Aehr and others that are helping your R journey. And there is a new site called rdevs.com
that might be worth a look. So we'll put that in the show notes, but thank you, Noveka, for sending it on and it's always great to see other, indexes in this space to help point our very enthusiastic, R users and data science to new tooling that they can use in their dev environment. Well, of environments, we're going to close-up this podcasting environment here so I can get to editing, but we thank you so much for joining us on this episode of R Weekly we and hope to have you back for our next episode of R Weekly Highlights
Our takes on the increasing removal of friction in good and not-so-good ways thanks to AI, a visual tour of hot air balloon accident data, and a case study of novel image recognition with the rplatipus package. Episode...
The flow of both data and analytical jobs just got smoother with mizu, opening the door for multiple agent harnesses in Positron, and an optimal statistical model tailored to complex experimental designs with a friendly...
What happens when that online data set isn't so easy to download anymore? We learn of two packages to ease the pain of obtaining these files and then some. Plus a visit to the visualization corner with a handy way to...
The celebration of rOpenSci's 15-year anniversary continues with the community sharing their perspectives, and a fun Shiny app that could be quite useful during celebrations on a weekend or the holidays. If you ever...
We learn about the variety of ways you can contribute back to the core of R, a roundup of the excellent choices available for package and environment reproducibility, and an amazing collection of Shiny applications and...