Copy of 2020-05-26-joshua-starmer.mp3

Joshua Starmer: [00:00:00] My story is you don't have to be the smartest kid in math class to be awesome at Data Science. I like math. But to be honest, it's not something I'm very good at. And I will be the first person to say math is hard. I don't just sit down there and read a math book and go, oh, I could prove that stuff. You know, I'm the guy flipping to the back of the book trying to find the answer to me, like, oh, my gosh, how they do that. That's genius! Because I struggle with it. I have to take it slow. And by taking it slow, that's how I know how to explain each step to everybody. When I break things down into small pieces, I'm not doing it for you or the audience. I'm doing it for myself.

Harpreet Sahota: [00:00:52] What's up, everyone? Welcome to another episode of the Artists of Data Science. Be sure to follow the show on Instagram at the Artists of Data Science and on Twitter at @ArtistsOfData. I'll be sharing awesome tips and wisdom on Data science, as well as clips from the show, join the Free Open Mastermind/Channel by going to Bitly.com/artistsofdatascience where I will keep you updated on biweekly OpenOffice hours that i will be hosting for the community. I'm your host Harpreet Sahota. Let's ride this beat out into another awesome episode. And don't forget to subscribe, rate and review the show.

Harpreet Sahota: [00:01:40] Our guest today is multitalented, native of Chapel Hill, North Carolina, who is not only a cellist, composer and singer, but a statistician and data scientist who has helped empower learners from all over the globe by breaking down complicated statistics and machine learning topics into small bite sized pieces that are easy to understand. He's earned a bachelor's degree in music theory and composition, as well as computer science from Oberlin College and PhD in computational bio stuff from the University of North Carolina at Chapel Hill. In addition to the creative work he's done composing soundtracks for several independent films, television commercials and modern dance productions, he served as a postdoctoral fellow and assistant professor at the University of North Carolina at Chapel Hill, where his work focused on developing novel statistics and visualization methods for high throughput sequencing technologies. However, you might better recognize him from his YouTube channel, where he's been creating videos since May 2011. As StatQuest creating educational and entertaining videos that demystify the complex concepts from statistics, machine learning and data science into small, simple, easy to understand steps as StatQuest. He's beloved by his audience of nearly two hundred seventy thousand subscribers and 12.6 million viewers for not dumbing down the material and instead build you the learner up so that you're smarter and have a better understanding of statistics and machine learning. So please help me welcoming our guest today with a thunderous DOUBLE BAM!, Dr. Joshua Starmer. Joshua, thank you so much for taking time out of your schedule to be here today. Man, I really, really appreciate having you here.

Joshua Starmer: [00:03:12] It's a pleasure to be here.

Harpreet Sahota: [00:03:13] Talk to us about your journey, how you first got interested in statistics and machine learning and kind of what drew you to the field.

Joshua Starmer: [00:03:20] It's a little embarrassing. So I first got interested in statistics as a graduate student. I wasn't a statistics major, but I had to take statistics classes and I thought the women in the statistics program were pretty cute. And I thought that if I wanted to get their attention, I had to be good at statistics. So I studied a lot. I mean that's kind of how it all started. I was in these classes and I was like, how do I get. How do I get these people attention?

Harpreet Sahota: [00:03:50] I wish I went to school where you went to school because people in statistics classes, they just all look like geeky guys like me.

Joshua Starmer: [00:04:01] No, I don't know. Maybe I'm actually attracted to geeky people. I don't know. But I just I mean, I just thought they were - they all knocked me out.

Harpreet Sahota: [00:04:09] Awesome, Man.

Joshua Starmer: [00:04:11] I loved those classes.

Harpreet Sahota: [00:04:12] So where do you see the field of Data science headed in the next two to five years?

Joshua Starmer: [00:04:18] I mean, I just see it becoming more and more important because we're only getting better at generating large datasets. I mean, the technology is just moving really fast. We're constantly in a almost kind of like a Wild West situation where we're generating data faster than we kind of know what to do with it. And so to me, that's a lot of fun of Data science is we get to be creative and come up with new ways because this Data is, no one's ever. I don't know. I just assume that next year they're gonna have a whole new way of generating crazy amounts of data, doing something new, and they're going to need us to come in and make sense of it. Some of that's gonna be using established statistics, some of that is just making stuff up as we go. I just see it being coming more and more important.

Harpreet Sahota: [00:05:04] Definitely. And I definitely make up stuff as I go at work from time to time and hopefully my boss isn't listening.

Harpreet Sahota: [00:05:12] So in this in this vision of the future where it's like the Wild West, bunch of Data coming in, like you mentioned, more than we know to do with, what do you think is gonna separate the great Data scientists from the really good ones?

Joshua Starmer: [00:05:23] I think, and this is true of any field. I feel like the great people are the ones that understand the main ideas and don't get lost in the details, because when you understand the main ideas, you can see a tool for what it truly is and what it's truly worth. And you don't get swept up in all the hype. And our field is full of hype and that's good and bad. You know it attracts people to the field. Smart people get sucked into it as well. And then the bad thing is, you have to kind of recognize despite the hype, tools are only good at doing certain things. And if you know the main ideas, you will know what tool is the right one for the right job. I think those are going to be the great Data scientists.

Harpreet Sahota: [00:06:05] Yeah, I think there is a, quote, something along the lines of the methods or the tools are many, but the principles are few.Right? [Inaudible] that the principles will really have that strong foundational grasp of the concept and have a real command.

Joshua Starmer: [00:06:20] Yeah, I couldn't agree more.

Harpreet Sahota: [00:06:22] Music theory, mixed with computer science and statistics. That's like really interesting combination to study in school. Can you first talk to us a bit about what music theory is, what a music theorist does. Maybe if you have done any research that ties music theory together with statistics. 

Joshua Starmer: [00:06:38] I can tell you what music theory is and what a music theorist does. And so what they do is; they, music theory is just a way to break down music into its components of harmony, rhythm and melody. So in some ways, what I used to do to music is what I now do with machine learning and statistics. I break them down into small components, which are really quite simple and easy to understand in and of themselves. So in some ways what I do now is just a continuation of what I was doing as an undergrad, where we would just look at a sheet of music and we'd say, well, what are the essential pieces that make this piece of music? Oh, lots of it make it special, but there are certain little key nuggets in there that really make that piece of music special. And now I just do that for algorithms since I look at the algorithm or what makes this algorithm special or what are the little pieces, how to actually break it into a little little things and understand a little parts and then put them back together. That's music theory. And I mean, I was really into music back then, machine learning was sort of like I mean, it was sort of there and artificial intelligence sort of there. But I don't even know if they offer statistics at the college I went to. It wasn't even really an option.

Harpreet Sahota: [00:07:47] I was watching this show. I think it is called American gods. And one of the episodes there's this like the dawn of A.I. or whatever. And one of the characters wrote like some type of program that statistically sampled music from, like Mozart or something like that and just created randomly, artificially creating music. And it just sounded really interesting and cool.

Joshua Starmer: [00:08:08] I did something like that as an undergrad. Yeah. I wrote a program that was basically based on sort of random number distributions or statistical distributions, and it would mean the effect was like really fancy sounding windchimes. And so the underlying algorithm was maybe more impressive than it needed to be. But that was an early experiment I did with sort of trying to use randomness or, you know, in a statistical sense. So it's not like pure random noise, but using randomness in a controlled setting to get kind of like a soothing kind of ambient music that you can just kind of play while you're doing your work rather than sort of like make a song that people have to listen to, is more just like what would be just cool to be like playing while I'm studying, you know, that kind of thing. 

Harpreet Sahota: [00:08:52] It is really interesting. I'm fascinated. Almost like a soothing white noise. Kind of.

Joshua Starmer: [00:08:57] Exactly, Yeah.

Harpreet Sahota: [00:08:59]  Do you think having a deep understanding of math has helped you be more creative as a musician or vice versa?

Joshua Starmer: [00:09:05] I definitely think creativity has helped me with math and also kind of, it's also a scary thing because I have to be careful not to, like, invent new things that don't actually exist in the mathematical realm. But I think it's more like the music kind of pours over. and I  think it all started with music to begin with and I just sort of cross in to machine learning for me. Music itself, though, when I'm playing my music or I'm writing my own music, it's strictly by feel. I don't try to think about it too much. And it's really my way of just sort of resetting my brain. I pick up my guitar, my ukulele, and I start playing. And my head just completely clears. And I mean, it's like a weird space where I'm just not thinking at all and I love that. With machine learning and algorithms, you have to be thinking all the time. There's no just like let's just feel this out and see what happens. But there's still like, there's definitely an element of discovery in both when I'm not thinking that I'm playing music. It's because I'm on an adventure and I'm looking for something new and I'm just going to see where I end up. And I, sometimes I end up really cool places. And that is sort of what happens with when I'm looking at new machine learning algorithms or I'm trying to understand that, I don't try to plan it out in advance how I'm going to learn or what I'm going to learn. I just dive in and I go let's see what happens. And sometimes things connect and sometimes they don't. And then if they don't, then I just put it down and I come back to it later and I don't try to force it.

Harpreet Sahota: [00:10:30] It sounds to me a lot like my process when I'm doing like exploratory data analysis, like everything you just described. This is pretty much OK when I first start a project and I first get presented a problem statement and a Dataset. Let us just slice dice it, summarize, aggregate, organized, feel it out, see what happens. Sounds pretty cool, man.

Harpreet Sahota: [00:10:54] What's up, artists? Be sure to join the free, open, Mastermind slack community by going to bitly.com/artistsofDatascience. It's a great environment for us to talk all things Data science, to learn together, to grow together. And I'll also keep you updated on the open biweekly office hours I will be hosting for our community. Check out the show on Instagram, @theartistofDatascience. Follow us on Twitter at @ArtistsOfData. Look forward to seeing you all there.

Harpreet Sahota: [00:11:22] I would love, love to get more into your music here, so if you could talk to talk to us about your music, and the work you're doing in that space. So what are some of the commercials and shows that feature your music.

Joshua Starmer: [00:11:34] My music has been used to sell, among other things, motorcycle helmets. So maybe you saw that motorcycle [Inaudible]. I actually don't remember what the brand was, but it was motorcycle helmets. It was kind of bizarre as like off road motors, like for like people doing like cycle. You know, they do this crazy jumps and things like that. But it was like I had this really soothing music, but it had a banjo line and somehow it like set the mood of like this dad buying his kid a new helmet for his birthday. I had this sense of nostalgia and that sold motorcycle helmets. But I've also done, I know I've done music for sort of like to encourage people to vote certain. A friend of mine who a filmmaker in Pennsylvania was working on a campaign there, and he used some of my music for that, but for films, I have done two like short kind of many films, like 30 minute long films. And one of them went to Cannes, which which was pretty cool. You know, I have had a little bit of experience doing that. And it's fun. I mean, it's I'll be honest, it's not something I probably want to do all the time. Actually, I love doing the statistics. And so here is what I learned and maybe I don't know, maybe this is a little rambling, but I will say some things I learned when I was doing music was I always wanted to innovate in the area of music. Maybe my music isn't actually that innovative in and of itself, but it's something I always wanted to do, whereas TV people and the movie people were always they'd always play me something and go, Well, we really want you to do is we want you to write something that pretty much sounds like this.

Joshua Starmer: [00:13:05] We just don't want to pay them because they're going to ask for a lot more money. So, well, we want you to do is write something that sounds a lot like it, just enough that we won't get so sued for copyright infringement. And I got that in the commercials and I got that in the films. And that was kind of like a rude awakening because I love just kind of innovating and creating my own thing. And surprisingly enough. And this took me a while to realize music really wasn't a good place for me to do my own thing. Oh, really good place for me to do. My own thing was it's, of all places statistics and machine learning where when I make a video, I just do what I want. I don't copy someone. I don't sit there and go, well, this is how they did it. So I'm going to do it just like them. But I'm going to change it just a little bit. No, I start from scratch. I go, how could I do this Completely different from anybody else? And that to me, where I kind of making that realization that my passion for kind of innovating and creating something new, that's really my own thing. I never saw that [Inaudible]. I grew up and I spent my whole life thinking that would be a music. And to me, the past couple years, I discovered that it was statistics and machine learning where I feel like I can express myself, who knew.

Harpreet Sahota: [00:14:19] That's very, very beautifully put, man, that got me right here. So you've got great music on on Spotify and on Bandcamp. Which album would you recommend that our audience go check out of yours.

Joshua Starmer: [00:14:37] I'm always, whatever it is, whatever my latest song or my latest album is, I'm always most excited about that because I feel like I've always learned from my mistakes or whatever. So my latest album, which is called Made for TV, it's on Bandcamp. Joshuastarmer.bandcamp. And I'm really proud of it. I mean, I'm proud of all my albums. So I started a [Inaudible] seven or eight years ago. For a New Year's resolution, I decided to write and record a new song every month. And I've been doing that for eighty five, eighty six, eighty seven months now. Yeah, I mean it was like a real struggle to make it through that first year. It was a real struggle to make it through that first month, like month one January, song number one almost didn't happen, but it did. And then I kind of just got momentum and and now I've been doing it for years and years and years. So every year I have a new album and my new album right now is called Made for TV. That's my baby.

Harpreet Sahota: [00:15:32] So,  what's the recording process like? Do you record everything yourself, do all the editing, all the mixing, all that stuff?

Joshua Starmer: [00:15:37] For the most part, I do it all myself. There was a phase where I had a band and I had a drummer and a bass player and a guitar player and we had a band and the bass player, had kids and then somebody else had something to do. And I kept reforming the band like every six months with new people. And it ended up just being a lot more logistically challenging because every time I reform the band, it will be like different people with different skills. And so we basically have to relearn the songs or rewrite them for a new ensemble every time. And now I just do everything myself. That's the way I started. It was everything myself. And then I went to band phase and now I'm kind of back to all myself, I do fantasize that one day I'll have band face again, because band is the band is really fun because it's cool to see what people bring to the music. It's cool to have, like, something in mind. But when you have a really good drummer and they bring magic, you know. And right now I'm not a great drummer, but I do my own beats and they're good. But they're not. They don't have that magic. So my fantasy is that the band will come back.

Harpreet Sahota: [00:16:43] That's it. That's a lot of work. That's impressive that you get all that done. And I can kind of relate to what you're saying about, you know, going from band to solo, because at the job I'm currently in, I am a one man Data science team, which is awesome. But like, you know, there's a lot of work I have to do by myself. And I kind of miss, like, my old company where I had, like, the most awesome team. And it's just like the magic of being able to work with other people and bounce ideas and creativity off one another to solve problems.

[00:17:12] Let's get into StatQuest now. Start with the genesis of StatQuest, what was the inspiration behind creating it?

Joshua Starmer: [00:17:19] So at the time I was working in a lab, a genetics lab with a bunch of wet scientists, and I was this Stat guy and they gave me their Data and I gave them their P values.

Joshua Starmer: [00:17:30] That was the deal. Well, I don't know. For some reason, I felt like you guys have to understand what I'm doing because I don't want this to be magic. I thought it would help them if they knew, like if they knew what was actually happening. So I started creating these sort of like Friday morning. I called it back then. I call it Stat chat. I Say, hey, we are going to have a little stat chat, you know, and I tried to explain what a T test was to them or I try to explain what a confidence interval to them or P value or all these like a basic fundamental statistics concepts. One day, just for fun, I decided to put it on on YouTube with the idea not that the world would be like, hey, stat chat, this is awesome. What it really was, was just a reference for the people I was working with because it's like I can do a video. And then when a new person comes to a lab and I have this question, I go, hey, just check out this video. Bam, you got it. So that was the original intention. And that's the way it was for a long time because nobody watched any of these videos except for like the three people I worked with. But then all of a sudden, you know, the way the Internet is, some people sort of watch and some of these people start subscribing. And then it kind of snowballed and grew well beyond my wildest dreams. I mean, they weren't even dreams back. It was not even a concept that people would watch this other than friends and family and coworkers. But, yeah, that's how it started.

Harpreet Sahota: [00:18:46] Yeah, that's crazy. And let me just cross 250 subscribers just recently, like in a couple of weeks ago. And since then, it's grown to nearly 270 by eighteen thousand in just that short amount of time.

Joshua Starmer: [00:18:58]  That's totally bonkers. It's totally insane.

Harpreet Sahota: [00:19:05] So I'm a mentor for a platform for up and coming data scientists and I often refer my mentees to your videos.

Joshua Starmer: [00:19:12] Oh wow!!

Harpreet Sahota: [00:19:13] To check it out, to learn. I've been referring to Statquest videos for a long time as well. And so all the contributions there, in that space has been amazing. I know they've helped millions of people, literally millions of people. I always like crazy.

Harpreet Sahota: [00:19:27] So what would you say is the mission of Statquest? What's one of the things that you want people to take with them when they come across your channel?

Joshua Starmer: [00:19:36] I mean, at least that people I used to work with, they'd be like, I cannot understand statistics. I'm a biologist, There's a reason why I took that class. I took as you know, I took a bare minimum and I never did it again because I'm a biologist or I'm something else. But what I really want people to take home is that anyone can understand these things. Ninety nine times out of 100, the only thing between them and understanding is fancy terminology and fancy notation. The concepts are simple. They're just hidden underneath intimidating lingo. That's what I want people understand that, that it's not them. It's not that they're not smart enough. It's not that they're not mathy enough. It's not. It's none of those things apply. What it is, is there's just fancy lingo and notation that's covering it up. All you need to do is have someone just push that stuff aside and you'll see that's the core nuggets of what data sciences are and you can understand it. Anyone can understand it. It's not a secret and it's not something that only math people can understand.

Harpreet Sahota: [00:20:40] When you're first confronted with this stuff, with the topics like it seems daunting with all these weird looking symbols and these abstract ideas, but the way you break it down, make it really concrete and digestible, it makes it super approachable, was there any like internal hesitation or fear with creating the content and then sharing it. How did you overcome it? Did you just go for it? I don't know if you're familiar with Seth Godin and Steven Pressfield, they have this concept called the resistance. So the resistance is that, it's an amygdalic response, right? It's your lizard brain telling you not to put your art out there for the world because you might be laughed at. You might not be appreciated by the audience, was there any of that. And if so, how did you overcome that?

Joshua Starmer: [00:21:26] There's always been a lot of fear and a lot of doubt. And to this day, all that stuff is there. A lot of it stems from the fact that my goal is to come up with something new and not just to repeat someone, what someone else said with like a minor twist. I've always thought that the reason why my co-workers did not understand statistics and their statistics class was whatever they were getting they didn't relate to it. So I've got to come up with something new. I can't repeat. I can't just say, well, I took that class too. And this is what they told me. I'm just going to tell it to you again. I needed to come up with something they could relate to. And when I do that, I'm coming up with a new way, or at least in my own opinion, it's a new way to present it. And that it's for me is very scary because I could be wrong. You know, it could be that my way of breaking it down, maybe I missed something. Maybe I didn't do, obviously, I didn't do the math right. But that's a scary thing. And on the one hand, it's great having a large audience. But that audience has also shaped me on the journey as well. Like, I can't anticipate everybody's questions or where people are going to get confused all the time. But over the years, people have watched my videos and they'll be like, this doesn't make any sense at all. And enough people will point to that one moment and I'll go, OK, that's something I need to anticipate next time, or that's the kind of thing that trips people up.

Joshua Starmer: [00:22:45] But it is very scary and intimidating. The good, the way I get over it, how do I get over this fear? Because when you have a lot of people watching your stuff and then you're wrong, it's very embarrassing. I don't know. It's kind of dark. It's not just embarrassing. It's kind of dark times. However, what I do is a crazy amount of research. And I do everything I can to check my work. I'll write programs to validate what I'm, my new or whatever my approach is. I'll take existing programs apart piece by piece, and I'll read them line by line until I understand everything that's going on. And what I like about that is it forces me to be more confident and more sure that I'm correct. And ultimately, because all that research, you know, I just take that. And then I poured it into the video and so everyone gets the benefit of my fear and doubt, which drives me to do all that research. It's probably a good thing that I'm a little nervous and a little freaking out because it pushes me just a little harder to make sure that what I'm talking about is correct and to do that research and not just kind of like knock stuff out, you know, each Statquest I do takes a long time and has a lot of effort goes into it. And that's why I just have to do a ton of research.

Harpreet Sahota: [00:23:53] This question might seem redundant after everything you just said but what's the most challenging part for you when it comes to creating content for the channel?

Joshua Starmer: [00:24:00] Well, I don't know if there's what's the most, but there's different parts. There's the part that says I can't do it. Like, I can't actually come up with a new way to explain this. Or I'll be honest, maybe I don't even understand the concept itself. Like, that's fear number one. Like when I was doing PCA., working on PCA, I didn't understand it, didn't make any sense to me. And that was one of the reasons why I was working on it to begin with. Because, like, I need to understand this for my work; for my co-workers, I gotta get this. But there's some doubt there, that's a scary thing to get over. But then there is, you know, the other hard thing, maybe it's harder, just the blank page, you know, and looking at that and go on, somehow I have to pull a rough draft out. If I can get that rough draft, I'm golden, just that blanks page. That is so scary. So that's the hardest part for me, is probably, you know, writing the first words or just starting to sketch stuff out. It's easy but it's so scary at the same time.

Harpreet Sahota: [00:24:56] I've been reading lots of Marcus Aurelius lately, The Meditation. He's got a line in there; The impediment to action advances action, what stands in the way becomes the way, the blank page is standing in the way and that becomes the way that you push through it.

Harpreet Sahota: [00:25:15] What's your personal favorite video from the archives?

Joshua Starmer: [00:25:17] My personal favorite is a series that I did on linear models. I've been thinking about linear models for years before I made those videos. It wasn't like I was spent 40 hours a week on it for years, but was just something I toyed around with. And I was like, there's gotta be a better way to explain this or has to be. And I remember also running with my friend Dominick, and I was like, I got it. I was so happy. And he was like, what? What just happened? I'm like, I figured out how to explain linear models. OK. Give it to me. And so I tried to explain it to him while we are jogging. And it was a total disaster. He's like, OK, I don't see what you're getting at. But I then spent the next maybe couple of months, maybe six months, even just sort of like starting to work on the figures and whatnot and how to illustrate things. I'm just really proud of that, that I got those videos done. And it's just something I've always thought could be explained in a easier, simpler way. And I believe I did it. So I'm really proud of those videos.

Harpreet Sahota: [00:26:16] That's awesome, man. Definitely have to go back and revisit that one and  I'll share those on the show notes as well. And it's interesting that, you know, you're on a run and then that epiphany happens. I think that's a common theme amongst most creative people is, you know, when they're out, away from the actual work, engage in some type of physical activity or meditation or what have you, that sort of thing actually come to light? So it's it's important, I think, to balance. Right, like not just be heads down into books or working all the time. Take some time away, go, the best of those insights come up. So would you say that is the absolute must for our statisticians and Data scientists to check out? It would be that series and videos?

Joshua Starmer: [00:26:55] I think so. If you're aspiring in either of those fields, I feel like linear models is basically, they form the heart, of a big part of what statistics and machine learning are. And if you can understand those, you well on your way to understanding all of it, because even other other methods will kind of fall back on regression techniques like a gradient boost uses a lot of regression techniques and a lot of regression terminology, if you understand or well versed in linear models, then, yeah, you are well on your way to understanding whatever other algorithms out there.

Harpreet Sahota: [00:27:33] Funny, you bring up gradient boost, actually. True fact here, I was using gradient boost and trees to solve something at work and I had to present it to my CEO and some salespeople, higher level salespeople literally. I went to the Statquest video on gradient boost and copy down everything you said and then re-digested it and presented that and they understood it.

Harpreet Sahota: [00:27:53] What would you say is the most misunderstood concept from statistics and machine learning.

Harpreet Sahota: [00:28:01] And why do you think that learners keep getting tripped up on it?

Joshua Starmer: [00:28:04] I don't really know what it is, but I will say it's something in a lot of people get confused about and it's real simple. It's that the probability that any, that a continuous thing like measuring height or measuring weight, that measuring any specific height or any specific weight, the probability of that occurring is zero. That blows a lot of people's minds. And they're like, what do you mean? Because like, how how could that part? I mean, clearly, if the average person weighs 150 pounds or whatever the weight is, how could the probability of weighing someone be 150 pounds be zero? And so the way I've starting to try to explain this is actually I might make a little mini Statquest. It just explains this one concept. Yeah, sure. Weighing one hundred. Someone who weighs one hundred pounds might seem like there's a high probability event, but what's the probability you'll weigh someone who weighs one hundred fifty point zero zero zero zero.

Joshua Starmer: [00:29:00] And then what's the probability? And that's probably a little smaller, right? Then what's the probability? You weigh someone who weighs hundred and fifty point zero zero zero zero zero zero zero zero zero zero zero. That's probably a little smaller. Right. And then you can just keep those zeros going all the way, you know, for a million miles. You just keep going and you and and then you're like, oh, yeah, the probability of having that precision, that much precision, that much accuracy. And then just nailing it ,one hundred and fifty on the head.

Joshua Starmer: [00:29:30] Yeah. That gets infinitely smaller the more precision we add to it. So I think that's the way I'm going to explain it. I think I'm going to make a little mini-, like two minute long Statquest just on that one concept, because I know a lot of people get stumbled on that. It sounds weird at first, but as soon as you start bringing in that precision, you're like, oh I get. Yeah, it's no big deal.

Harpreet Sahota: [00:29:57] Are you an aspiring Data scientist struggling to break into the field, well then check out dsdj.co/artists to reserve your spot for a free informational webinar on how you can break into the field. This would be filled with amazing tips that are specifically designed to help you land your first job. Check it out: dsdj.co/artists.

Harpreet Sahota: [00:30:23] Is it important to learn all of the formulas and equations? Even though we have that software that does the work?

Joshua Starmer: [00:30:31] I know one formula. I know the formula for the standard error and look how far I've gotten. I don't think it's important at all. I know I probably a ton of people would disagree with me, but I'm one of those people where Google is my long term memory and I prefer to store just the essential concepts, the ideas.

Joshua Starmer: [00:30:52] Give me the ballpark. If I need details, I can always research those later. So I'm a big sort of like big concept person. Focus on those equations, though.

Joshua Starmer: [00:31:03] They're there when you need them. I've never been like I wish I knew this equation off the top of my head. I've never been in that situation. And I don't know many equations off the top of my head. I mean, actually, I was just looking at the normal distribution equation yesterday, so maybe I could tell it to you off the top of my head. But if I had just looked yesterday, I probably like. I don't know

Harpreet Sahota: [00:31:27] So once upon a time, I was an actuary. And there was a series of exams for that and you'd have to memorize all of the formulas. So if I recall correctly, I think the normal distribution formula is one over two Pi Sigma squared E to the negative X minus mu squared over I don't know something like that.

Joshua Starmer: [00:31:53] So you've got a lot farther than I would have, that's impressive.

Harpreet Sahota: [00:31:57] These things haunt me

Joshua Starmer: [00:31:57] Amazing to me. How do you think that would be useful in your daily life?

Harpreet Sahota: [00:32:03] The entirety of exam one is just theoretical probabilities, foundational probability, and it's having you're taking the exams and the only thing you're allowed to bring into the exam is a T.I Calculator and not like a T.I 83.

Harpreet Sahota: [00:32:20] But like one, just a regular T.I Calculator and you have to do everything by memory. It was brutal man. It's one of the reasons that I had left that field. I couldn't, after four exams, I just couldn't do anymore. 

Joshua Starmer: [00:32:30] That's not for me.

Harpreet Sahota: [00:32:37] So I was looking at your, at the facts on your website. Question one. Can you recommend a good book for learning statistics? I thought the answer was hilarious because I knew it to be true for my own reasons. But I was wondering if you could elaborate. First, tell us what your answer was and maybe tell us why. Why is it so hard to find a good book for learning statistics?

Joshua Starmer: [00:32:59] And the answer is no. I mean, I'll be honest, I have not read a billion statistics textbooks. I've read a handful of them. And some of them kind of speak to me a little bit, but most of them don't speak to me at all. I mean, my explanation is there's not a lot of innovation in teaching. And I've always thought that if you want to educate someone, you have to connect with them and you have to relate with them and you have to see the material from their perspective. It's not yours, not the way you were taught. You know, like, oh, I was taught this way. So I'm going to teach this way, no. You have to say, what are your experiences? How do you see this problem through your eyes? How can we see the solution to that problem? And I just, I'll be honest, I don't see a lot of educators actually doing that. I think it's really important to understand what it's like to be scared of equations, to understand what it's like to look at Greek. And I apologize to all the Greek people in the world. But for a non Greek person to look at those Greek characters and go, I don't even know what that is. That's a weird squiggle. You know, that's a,there's a lot of fear and a lot of, like, confusion that can often be replaced by just using a word, instead of symbols to represent something. It's all about trying to communicate.

Joshua Starmer: [00:34:11] And I just don't think a lot of teachers are thinking about communicating. They're thinking about the material. Sure. They're thinking about what the important concepts are or they're not thinking about communicating. And that's the whole point of a textbook, is to communicate a concept.

Harpreet Sahota: [00:34:27] So I wanna get your perspective on; if you consider Data science machine learning to be an art or purely a science. And and why?

Joshua Starmer: [00:34:36] It's a little bit of both. I mean a lot of data science is grounded in solid statistical theory, but not all of it is, to be honest. I get it. There is an element of the Wild West of Data science is just a few steps ahead of solid statistical grounding because the data is just coming at us too fast. I mean, has anyone done like a full statistical underpinning analysis of XGBoosts? I don't think so. I know how the equations were derived. But has anyone proved that it's like asymptotically normal or you know what I'm saying? Like, there's all this stuff that kind of comes in later often, whereas there is all these pioneers that are just of like chipping away with with whatever seems to work. There's a lot of that in Data science and machine learning where we're like, well, it works. I mean, not all of it, but but there is a lot of it. So there is an art form there. But there's a science because there is a lot of statistical theory and a lot of the older techniques are really well grounded.

Joshua Starmer: [00:35:36] Just because they're older doesn't make them any less useful. I mean, the hammer and the screwdriver have been around forever. I still use them all the time. I don't always need to use my you know, rechargeable battery, circular saw whatever, I can use a hammer and a screwdriver. So, age isn't really a factor. And so there's a lot of statistical there, plus when we're talking about art, let's talk about drawing graphs and like presenting data and making it look beautiful, but also interpretable. I mean, you can go both ways there. You can get artsy in ways that, like, makes your graph and Data inscrutable or you can make it clear and crystal and that's it. I feel like that's an art form, true color choice. It's something that's all important. It seems like a little thing, but it's a big deal, especially when you're around somebody who knows what they're doing and you're like, dang, that is beautiful data. And that's a that's an art form right there.

Harpreet Sahota: [00:36:25] So what role do you see being creative and curious plays in being successful as a data scientist?

Harpreet Sahota: [00:36:31] And how could someone who doesn't see themselves as creative actually be creative?

Joshua Starmer: [00:36:38] Being creative, I believe is critical to Data science, is like it's fundamental because what do we do? We create insights. We create new knowledge out of a pile of Data. So that's creation. And that's a very creative act. And that's what we do every day. That's what we're always asking to do. It may not seem very creative because we're just like applying an algorithm or sitting at a computer but we are generating new knowledge and new insights that never existed before. And I believe as people, you know, if you don't think you're creative, okay, that's fine. But I think you should think again, because I feel like people are, by their very nature, creative. We create things all the time. We create children, for example, they're not copies, but they're new and unique. Not everyone, but most everyone can do that. I mean, it's not a big deal. I mean, some people do it by accident. I mean, so and that's being creative. There's all kinds of ways to be creative that you may not realize are very creative. You're creating new stuff. So if you don't think you can create something new. Think again. Can look around because chances are you are. And Data science is just another way to do that.

Harpreet Sahota: [00:37:50] Very beautifully put, speaking of creating children, I just, my wife and I just had our first baby a few weeks ago.

Joshua Starmer: [00:37:54] Oh wow, that to me is the ultimate creativity.

Joshua Starmer: [00:37:58] You know, I feel like that's the top of what humanity can do. Yeah.

Harpreet Sahota: [00:38:04] Beautifully put.

Harpreet Sahota: [00:38:05] So what would you say are the similarities and differences in the creative process for, let's say, writing a research publication, composing music or creating youtube video?

Joshua Starmer: [00:38:16] One thing is similar as they all start with a scary blank page. They all. That's just, when you're creating something. You start with nothing and you gotta build something. You know that. And that's always super scary. And so they all have that in common and they all have tricks and ways to kind of motivate yourself to get over that. They all have to have some sort of inspiration that's going to keep you from blank page to something. The big difference for me. For all those things is knowing when I'm done. And music being the most ambiguous of them all. Whereas like when I do research publications, you kind of know what you want to do. You got to cover these points. You got to tell a story in a certain way. You've got your Data, you want to hit on those figures. When I'm doing Statquest, I know what the concept is. I need to explain. You know, I may not always know what the best way to go about it, but I can start chipping away. And I and I do recognize once the concept is there. Like, yep, that's solid. But with music, it's like, what do you mean we're done? I mean, I could I could throw everything away and like start over with the piano or I could add a drums or you know, it's like there's less that says when you're done. In some ways that makes music a little scarier because you just have to decide you are done. Like, well, it's I guess that is the song. I guess music's scary at the beginning, fun in the middle and then scary at the end. Whereas everything else is sort of scary in the beginning and then just sort of fun for the rest of the way.

Harpreet Sahota: [00:39:38] Last question before we go into lightning round. OK. And that is, what's the one thing you want people to learn from your story?

Joshua Starmer: [00:39:46] I think my story is you don't have to be the smartest kid in math class to be awesome at Data Science. I like math, but to be honest, it's not something I'm very good at. And I will be the first person to say math is hard. I don't just sit down there and read a math book and go oh yeah, I could prove that stuff. You know, I'm the guy flipping to the back of the book trying to find the answer to me, like, oh, my gosh, how they do that. That's genius. But because I struggle with it, I have to take it slow. And by taking it slow, that's how I know how to explain each step to everybody. When I break things down in the small pieces, I'm not doing it for you or the audience. I'm doing it for myself. And I think a lot of people look at me and then go, oh, you must be good at that stuff. But it's actually the opposite. It's because I struggle that I'm good at explaining it. And that's why I think anyone can learn this stuff. I feel like if I can do it, so can you. That's what I think people should take home, that it's never out of their reach.

Harpreet Sahota: [00:40:47] I love it man, very well put. So let's go ahead and jump into the lightning round. How did the catchphrase "bam" started?

Joshua Starmer: [00:40:56] I really don't know. I just started saying it a lot and I still say it a lot. And if you're hanging out with me, I will probably say it once every couple of minutes. I just say, bam! And then I was realized I needed to add emphasis because some things were even cooler than bam!. So I started saying "double bam" and "triple bam" was just the next logical step. Yeah, I don't know. I mean, I think I was talking to my lab mates and I was like, check it out, guys, this is how statistics works, bam!

Joshua Starmer: [00:41:26] And there it was; the "bam". Let it out of the cage.

Harpreet Sahota: [00:41:30] So what do you see Stat Quest becoming in the next two to four years?

Joshua Starmer: [00:41:36] I would love ultimately for Stat Quest to become more than just me. So right now it's a one man operation. I'd love to grow it. Because I feel like there's so much more material we could cover. I can only get so far.

Joshua Starmer: [00:41:49] And it would be fun to kind of work with like minded people and just cover a whole lot more. So that's sort of like a fantasy. Another fantasy I have, which I'm getting closer to, is also doing like a diploma, like a stat quest diploma thing. I think that would be fun to do. And I've been kind of chipping away at it from like random angles for a long time. But I think it's finally starting to take shape as a legitimate curriculum so that may happen in the next two to four years. I'm looking forward to both those things.

Harpreet Sahota: [00:42:28] So apart from your own music, what are your top three favorite musical artists?

Joshua Starmer: [00:42:36] So this is kind of funny. I don't actually listen to very much music at all. I mean, I know that sounds crazy, but it's actually true. I rarely, if ever, listen to music. That said, about 10 years ago, I was really into Andrew Bird. I thought he was fantastic. But now mostly I just see him as a I'm a big fan of Beethoven. I love Beethoven. I listened to like 15 minutes of classical music every morning. If I do listen to music, that's what I listen to. Not that I'm some, like, weird person and I am a tower off, but I mean, people are playing music around me all the time. And I always hear, meet new music. And I've got to as a musician, I've got a lot of musician friends that are always playing me their music.

Joshua Starmer: [00:43:17] I'm around music, but I don't actually pursue listening to music.

Joshua Starmer: [00:43:22] Just I know. Weird.

Harpreet Sahota: [00:43:24] I was going to ask you what your favorite song was.

Joshua Starmer: [00:43:26] I have no idea.

Joshua Starmer: [00:43:31] But I do have songs that I like. There's a song that I recently heard by a relatively contemporary band that I just think is fantastic. And it's been stuck in my head for the past month and a half. And I've only heard it once, but it's just been a lot in there. It's a song called Fortune by a band called Wye Oak. I had a friend who engineered, so basically they set up the microphones. They made sure everything was sounding good and they recorded everything. And I know the people that mixed it. And so I heard it through that way. And it's just been like, wow, that's a crazy good song. So even though I don't really listen to music,I guess I have a song stuck in my head.

Harpreet Sahota: [00:44:11] So what's the number one book, fiction, non-fiction or both that you'd recommend our audience read and your most impactful takeaway from it.

Joshua Starmer: [00:44:20] I wish I can recommend something deep. But what I've been rereading all of Neal Stephenson's books right now and I don't. Have you ever seen it, these things are monsters? There's like a thousand page tomes that he writes in the guise of like an action thriller. But there's actually a good amount of philosophy lodged in there. And I just love the combination of like action packed thriller and philosophy does that. And I could just watch. I could just read Neil Stephenson books over and over and over again. There's a sort of so rich and there's the characters. I always see new things in them every time. I just love them. I mean, they're light and fun, but also deep and rich.

Harpreet Sahota: [00:45:00] I will check that out.

Harpreet Sahota: [00:45:02] So if we could somehow get a magical telephone that allowed you to contact 18 year old Joshua, what would you tell him? First, give us some context. Eighteen year old Joshua. What was he up to? Where was he at? What would you tell him?

Joshua Starmer: [00:45:17] So at 18 years old, I was determined to become a professional cello player. That was my goal in life. And and in addition to that, I saw myself as maybe going into film scores. I was I mean, back then, I was all. I was writing music back then.

Joshua Starmer: [00:45:36] Either way, I feel like I've got to put an asterisk to this, people are familiar with Stat quest theme songs. That's not what I'm talking about when I'm talking about my own music. The stat quest theme songs are intentionally silly, unrehearsed and just spur of the moment spasms of fun. Whereas like I am, when it comes to me writing music, I'm very serious about it. I've always been very serious about it. So don't get confused by me singing in a terrible falsetto with a banjo, that's different, I had to make that distinction. They were like, holy smokes, what is this guy talking about? But when I was 18, I really just wanted to be a cello, I loved playing the cello more than anything. I just had a real strong passion for it. I wasn't great at. I wasn't very good at it, but I just loved playing it. And I loved music and I loved writing music. I loved film scores.

Joshua Starmer: [00:46:30] I just had this fantasy that those would be things that I'd be doing at this stage in my life right now. If I had to go back and and tell myself what actually happened, I don't think I would believe it.

Joshua Starmer: [00:46:47] I mean, to say that, you know, you actually had those opportunities to you know, you played cello professionally. You did write some soundtrack music and you didn't enjoy as much as you thought you'd enjoy. But what you really thought would be fun. What you know, what ended up being, like, super fun and super passionate, statistics. I mean, what 18 year old is going to like. Yeah, right.

Joshua Starmer: [00:47:09] You know, no way. No one's going to believe that.

Harpreet Sahota: [00:47:13] Do you have a favorite film score?

Joshua Starmer: [00:47:15] Oh, no, I haven't listened to it forever. But when I was a kid, when I was 18. Peter Gabriel scored The Last Temptation of Christ. It's a Martin Scorcese movie that I have never seen. I just remember there were protests like crazy but the soundtrack blew my mind. It was one of the most beautiful things I'd ever heard. And I think I listened to it every day for like five years. I could not stop listening to it. That was sort of a big transformative thing. I was like, I want to make music like this. This is pure beauty. 

Harpreet Sahota: [00:47:53] So what's the best advice you've ever received?

Joshua Starmer: [00:47:56] I've gotten to two, two things. My father always said do something you're passionate about. He also said focus on the main idea. So those are two good things from my pop. My boss always said; do the most important thing you can do. And that advice, when you're younger. When I was younger, I didn't know what the most important thing I could do was. And so that advice isn't super helpful. I mean, you can try to search for the most important thing you can do. But for me, I just stumbled over it. And for me, that was Stat quest. About a year ago, it just dawned on me that Stat quest was the most important thing I could be spending my time on, because that's, if I'm going to make a mark on this earth, that's how I'm going to do it. I'm going to in my own little world, in my own little space, I'm going to try to revolutionize how this material is being taught. And I think I can do it. And it may not reach that many people. But I think that's my gift and that's my skill. And that's something I'm really good at, that I think is special about me. And that's what I need to be doing. I need to be doing Stat Quest. I need to be explaining statistics, I need to be explaining machine learning. I need to spend as much time doing it as I can, because that's how I'm going to. That's how I'm going to help. I can do anything to help make the world a better place. That's how I'll do it.

Harpreet Sahota: [00:49:24] And that really resonates with me, man. It really hits me. I kind of feel the same way about this podcast. Yeah. I mean, I'm hoping that by sharing other people's stories, sharing other people's journeys, that people realize that it's not, you know, peak to peak, that people go through ups and downs and people have to actually learn this thing. You're not just fresh out the womb a statistician, right? You put in the work, you put in the effort, you put in persistence over a period of time and then you develop mastery. And, you know, I hope that's what my audiences is gaining from every interview that I'm doing. But that really resonates with me.

Harpreet Sahota: [00:50:05] So what motivates you?

Joshua Starmer: [00:50:08] Well, I talked about this a little bit. Is the big motivation, or least right now is I've only you know, I won't be here forever. I've got to get this done now.

Joshua Starmer: [00:50:21] So that's a big, big motivator, as I know. I know. It's something I'm good at. I know I can do it. And I know that's the most important thing I can be doing. So that's a big motivator.

Harpreet Sahota: [00:50:34] How can people connect with you? Where can they find you online?

Joshua Starmer: [00:50:38] Well, one way, it is an easy way is to comment on a video if you can. But you can also contact me through my website, statquest.org. and I'm also on Twitter and LinkedIn. So, yeah, there's lots of ways to get in touch with me. I will say I don't always respond to everything I get, but I do read everything; good, bad, love me, hate me. I will read anything that gets sent my way. I don't. Like I said, I don't always respond. You know, someone, A lot of people ask me like, hey, can you do a stat quest on on a topic I've never even heard of before? And I actually like those emails, you know, because next thing you know, I'm like i am googling it. Like, what is this? Oh, this is cool. And I'm getting into it, but I don't always respond to those because I don't, I can't come up with an answer to that. It's like I could. And if I had a lot of time I would get to it. But I don't know. I can't make any promises. So I'm like rather than make myself feel uncomfortable, I'm not going to respond. But I do read everything.

Harpreet Sahota: [00:51:41] So how do you not let the negative comments affect you?

Joshua Starmer: [00:51:44] Like, oh, that takes practice. That was really hard to do very early on. I wasn't expecting it. Like, I remember the very first really negative comment I got took me real bad because, I mean, let's face it. What am I trying to accomplish? All I'm trying to do is help you understand your homework. Not like, not hurting anybody or not. You know, I went into a really naive, I mean, the Internet is a big place and it's fun to kind of build up a channel and build up a reputation. But, you know, imagine this normal curve and we've got ninety five percent or even ninety seven percent of the people are on the good side. We're here. There's that 2.5 percent that once you start sampling large proportions of the overall population, sooner or later going to hit those people. And what do I do? I mean, the first time it hit me really hard. Now I'm better at. I don't know, I don't even know how to say it. Like, I don't, I try not to take it personally. I will say early on when when I got to seat, I talked to some friends about it. My friend Brian Risk, he said, well, that guy obviously has no life. And I was like, oh, yeah, that's probably true. You know, the way he just was like he didn't even think about it. He was just like, clearly he has no life. I mean, what you're doing is great.

Joshua Starmer: [00:53:01] And so now it's like, you know, when I see the you know, every now you just get someone who really hates my intro song or something like that. And I just hear Brian on my shoulder. He goes, don't worry about it, you know, whatever, you know. And I'm getting better at it over time. But it was a big shock early on, when, you know, it's not like I got a lot and it's not like, it's literally, it's probably like one in a thousand comments. I mean, once every three or four months, you know, when you've got all twelve million people watching your videos or something, like there's some crazy people out there and there's, but everyone else is great. But when you get it, I don't remember the first time I was like, oh my God, how can they say that about me? But now I'm getting a little bit better and have a better perspective of like, don't forget about that ninety seven point five percent of the vast majority of people that are good people. And it's just the occasional weirdo on the Internet. And that's just that's the nature of the Internet, you know. So if anyone who gets a large enough audience, you're inevitably going to run into all these people, because this is what they do for their hobby I guess, that's what they think is gonna be the way they mark the earth.

Harpreet Sahota: [00:54:09] It's a horrible imprint to leave. You're making an awesome an awesome imprint. You contribute so much to the community. Thank you so much. And I just thank you for taking time out of your schedule, to be here on the show. I really appreciate it man. Thank you.

Joshua Starmer: [00:54:25] Yeah. Thank you very much. It is very flattering to have you hosted me. It's always been a dream to be on a podcast. And this is my first. I love it. Thank you.