Hi everybody, welcome to reproducible research. The purpose of this lecture is to introduce you to the idea of reproducibility and to explain why I think it's really important. Even if you dont think of yourself as a researcher or as conducting research, if you analyze data at all, it's critically important that you understand the ideas of reproducibility and the problems that it introduces. So I want to get that across to everybody right off the bat so you can understand it and see why it's so important. Now, in case you're not familiar with the idea already, I thought I'd introduce some of the issues that are raised by reproducibility by drawing an analogy from another area, and that's music. So before we go on any further, just take a quick listen to this excerpt that I'll play for you. [MUSIC] Okay, so that was a song called Code Monkey. It was written and performed by Jonathon Colton, and I just played you the first minute or so of the song that got you through the first verse and the chorus. So you might want to just rewind and just maybe listen to that excerpt one more time just to kind of get it in your head, it may get stuck in your head, so I apologize in advance. But, think about that song and what you already know about it, all right? So, you know what it sounds like. You have a sense of who wrote it, who performed it. You know that it starts with a guitar in the beginning. If you have a pretty good ear, you know that it's in F major. And then later on another guitar joins as a drum kit that plays along. And of course, he's singing along. He's singing the words. And if you can listen carefully and you understand a language, you could even tell what words he saying. And so, and that's pretty much the whole song. And of course, I didn't play the whole thing for you, but if I had, you could have decoded a lot of this information just from the recording, okay? So most popular songs like this are not very complicated. They don't have a lot of instruments or a lot of voices, and they typically range from two to four minutes. And if you're a good musician and you can listen very carefully, it's not that hard to kind of play the song, maybe on your own guitar or whatever, after you've heard it, okay? Now, okay. Now, let's take a listen to this piece now. [MUSIC] Okay, so in case you didn't recognize it, that was the very beginning of the Symphony Number Eight, by Gustav Mahler. It was performed by the Chicago Symphony Orchestra, under the baton of Sir Georg Solti. And so this might be the polar opposite of the Code Monkey song that I just played for you. This symphony is sometimes called the Symphony of 1000, because just of the sheer number of people that have to be on stage to perform this piece. You need an entire symphony orchestra. You need an entire full chorus to sing. And so there's a lot of complex moving parts needed to perform this piece, and yet it gets performed all the time, and everyone recognizes the piece when they hear it, because it's more or less the same thing. Now how does that happen, right? So how is it that orchestras and choruses all over the world can play this enormously complex piece and it always seems to come off more or less just the same thing, okay? And similarly, how is it that you can listen to Johnathan Colton's song and you can hear the music, and then somehow, if you are trained musician, you can pick it up and maybe play it for yourself, okay? This is really about reproducibility, right? So in music, one of the nice things about music in general is that when you hear the performance, regardless of whether it's a simple popular song or an enormous symphony, you get all the information that you need. And now depending on the complexity of the music, you may get more or less information, because the brain is only able to process so much information at a time. But for something that's very complicated, like a symphony, we actually have a way to write down the instructions to give to the performers to tell them how to play each thing. And so it actually, Mahler is a great example of this because he himself was a conductor. And he knew that as a conductor it's often difficult to look at a composer's music and understand what is it exactly that the composer wanted here? And so when he wrote his music, he put instructions on every single inch of that piece of music, because that way the conductor who is interpreting and the performers who are interpreting the music can understand, oh, that's what the composer wanted here. So for Mahler's Symphony Number Eight what we have is the score. So the score is basically a book that lists every single part of every instrument, and what they need to play, and what instructions they have. And so the conductor can look at the score, and say, okay, I know what the violin's playing at this time and I know what the flute is playing, and I know what the chorus is singing at this time. And that way the music can be coordinated and all performed, and synchronized. You can have a score for a popular song too, so Code Monkey could have a score, there'd be guitar lines, there'd be drum lines. It's typically not written in the same way that you would write a symphony, but there is a notation that is used sometimes to communicate how a piece of music should be performed. And so what we're talking about in this course, Reproducible Research, is basically, how do you develop the score for data analysis so that you can communicate to someone what was done, and if they want to reproduce the work, how to do it, okay? Now the fundamental problem in data analysis is that we don't really have an agreed upon notation system for communicating data analysis. And so, everyone does it a different way, and in aggregate, it's kind of a mess. So some people will just describe in words what was done, and in some cases, this is suficient. This is sufficient but in many cases, it's not. Some people will provide the computer code and the data and everything that you need, and sometimes that's good. But sometimes it's enormously complex and it's difficult to sort through. And so there are a variety of ways that you can communicate data analysis, but we just haven't agreed upon what is a way that we can all look, that is sufficient for everybody, or more or less sufficient for everybody. So, what we're gonna focus on in this course, is how to communicate data analysis, using code by writing documents that are very dynamic, and by sharing data so that other people can reproduce the work that you're doing. So this is important for all data analysis. It's not just about doing research per se. And because if you wanna communicate what you've done to other people, you need to be able to give them the material so that they can perform, so to speak, the analysis themselves.
