1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
Please read the line below:

TFP 2016 Updated!

R and other Analysis tools

Discussion in 'Tilted Gear' started by genuinemommy, Aug 30, 2016.

  1. genuinemommy

    genuinemommy Moderator Staff Member

    Do you use R?

    What else have you used?

    Here is a place to discuss R.
    Or to rant about it.

    This is the second time I have messed around with this. The first time was to do some fancy statistics. I ended up giving up when I learned that there was an expert statistician available to help out on-campus, and just handed him my data set.

    Now I'm here again... different application, same coding mess.
    I hate R.
    I hate code.
    I don't understand it. Every tutorial I have found shows me I've jumped in way over my head. I have a lot to learn in a short amount of time and I'm not happy. Supposedly there's a bunch of support out there for the R environment, but... I don't speak informatics. yet? Grr. This is not the direction I wanted to go.
     
    Last edited: Aug 30, 2016
  2. I've only ever used SPSS for quantitative analysis.... R does not sound fun.
     
    genuinemommy likes this.
  3. fflowley

    fflowley Don't just do something, stand there! Donor

    It sounds like my worst nightmare.
    All the clinical trials we do have dedicated stats nerds.
    Sorry.
     
    genuinemommy likes this.
  4. Japchae

    Japchae Very Tilted

    I cannot participate in this thread due to post-traumatic statistics disorder. And without using an astronomical amount of expletives. I. Hate. R.
     
    ZombieSquirrel and genuinemommy like this.
  5. rogue49

    rogue49 Tech Kung Fu Artist Staff Member Donor

    Location:
    Baltimore/DC
    Just make sure you use optimized SQL to extract your data from very large sets
    ...otherwise, it will take forever. (had to teach the data scientists this...)
     
    genuinemommy likes this.
  6. Shadowex3

    Shadowex3 Very Tilted

    My graduate cohort was taught R instead of SPSS and it was at the time an extremely shortsighted decision on the department's part, enough of us raised a stink with the graduate coordinator that teaching R to grad students as their first introduction to quant methods got shitcanned. That was one of the things that earned me in particular some bad blood within the department. I've since picked up some SPSS from my thesis chair and come to appreciate R more myself.

    I'd say each has its strengths. SPSS is probably one of the best dataset managers I've seen so far; it handles huge sets like the World Values Survey gracefully, lays it all out in spreadsheet form for you, and makes some operations like reverse-coding and some kinds of variable recoding very easy and painless. On the other hand for actually doing some things like complicated subsetting, looped operations, and splitting analyses up R is extremely powerful and it can spit out production quality graphics almost trivially easily once you know the right functions (hint: ggplot2)

    I think I really warmed up to it after I spent a little over a year helping teach undergrads that didn't even know what a directory was everything up to linear regression in R. Admittedly I described it as "fifty shades of R" (no connection to the same-named site) on more than one occasion, but it's not impossible to learn. Painful, but not impossible.

    It's basically a lot like someone took python, made it slightly less of a coding language and slightly more of a command-line based environment like STATA, and told it to index from 1 instead of 0. Some things you want to use raw code for, a lot of things you can use precanned functions for, and using an environment like Rstudio helps enormously compared to the stock GUI.

    The learning curve isn't exactly linear. When you start out most people feel like they're trying to hold too many different things in their head that they don't understand all at once. Then there's a point where you just "get it" and suddenly everything you've been struggling with just seems to make sense and you move on to bashing your head against the next wall.

    If you're still struggling with the R environment itself (the language, how it works, etc) downloading Rstudio and going through the interactive lessons in the Swirl package will really help. If you're comfortable with how R works but don't know how to do the things you want to do with it I'd move on to some of the O'Reilly books. I've got a few of them in PDF I could send you, along with the graphics cookbook and a troubleshooting book called "The R Inferno".
     
    genuinemommy and rogue49 like this.
  7. rogue49

    rogue49 Tech Kung Fu Artist Staff Member Donor

    Location:
    Baltimore/DC
    @Shadowex3 ...you just took it to the next level :cool:
     
    genuinemommy likes this.
  8. Shadowex3

    Shadowex3 Very Tilted

    Well like I said, I helped teach it for 2 semesters and kept dropping by just to help voluntarily for something like a year after my assistantship ended. I've been contemplating writing an undergrad level text on it myself, something that starts out explaining basic computer literacy and moves on through "everything is an object, everything that happens is a function call" R environment theory to actual basic statistics (crosstabs, mean comparison, linear regression, some graphics, basic function writing and loops, etc).

    There's stuff like Swirl and books like O'Reilly but I haven't seen anything that kinda bridges the gap between extreme "2+2=4 see it's a calculator too" basics and straight up coding tomes.
     
    rogue49 and genuinemommy like this.
  9. genuinemommy

    genuinemommy Moderator Staff Member

    This book needs to exist. Please make it.
     
    rogue49 likes this.
  10. While looking over the new requirements for my degree (luckily it won't throw me off too much) I learned there is a new stats course we have to take....It would have been a nice refresher before the Quant class I took. I might try to get out of it...I also might take it for fun since I actually kind of like Stats. I do think the class was added because so many of us were out of practice by the time we took Quant analysis. Our prof probably thought we were better prepared than we really were. I did well in the class, but could have done better.
     
    genuinemommy likes this.
  11. Shadowex3

    Shadowex3 Very Tilted

    Tbh I think stats and quant methods need to be two different classes. One should be focused on teaching theory and statistics, the other should be entirely focused on teaching you how to do all the different things you can do with your department's preferred stats software.
     
    genuinemommy likes this.
  12. genuinemommy

    genuinemommy Moderator Staff Member

    After 3 hours of skype chat with @shadowex3 I feel way less intimidated. If any of y'all need to understand R, he's your guy...
     
    Chris Noyb, redravin and rogue49 like this.