Amelia McNamara | Working with categorical data in R without losing your mind | RStudio (2019)

[@muhammedhadedy4570]

Excellent presentation Prof. Amelia. I learnt a lot from you today, thank you so much for your work.

[@andremunoz3939]

What is the solution to the test/training selection error?

[@BRichard312]

I was only able to view about half of this video before I had to turn it off. This presentation relies on such poorly presented and obsolete ideas that I cannot even understand how this character is able to continue teaching statistics. Amelia, you have an OBLIGATION as an "assistant professor" of statistics to maintain and update your knowledge of R. By definition, you are not doing that, and it shows in this completely obsolete presentation you delivered in 2019. The solution to factor problems in R is a function found in the admisc package called frelevel. It completely resolves the factor issues you presented in your talk. Secondly, if you want to work with data of class factor literals as well as levels, then you create two separate vectors for each field you are using when working with them. One field represents the factor data, and a corresponding vector represents factor levels. While it may be argued that doing this increases the scope of the data set in an unwieldy way, it will prevent the problems you outlined when working with factor data in R. And no, the excuse of "well, this presentation was given in 2019 so the technology was not available then" is BS. The frelevel function has been available since April 2019. If you were maintaining your R skill set, you would have known that. You then reference linear regression - another obsolete tool used in statistics. Are you kidding me? I cannot believe linear and logistic regression are still being taught in statistics classes today. The only time LR should be used is to determine measures of association - NOT prediction. There are far superior capabilities in R that address prediction than Linear or Logistic Regression, not the least of which include but are not limited to recursive partitioning, neural networks, auto-encoders, Naive Bayes Classifiers, SVM, KNN, and k-means clustering. Your teaching is anachronistic at best here. You need to review CRAN for the latest thinking on what is happening in how new ideas in statistics are being applied today because what you are teaching is obsolete. God help us if you are still, in 2021, teaching this same nonsense from this 2019 presentation. I wouldn't be surprised if you were. If you are, you are doing all of your students a profoundly egregious disservice. Your students are paying for grossly misinformed data. And, quite frankly, if the functions you are referencing are obsolete, it is highly likely that the process models you are teaching that are supporting these functions are likewise obsolete. If I were a student in your class, I would move to have you removed as a professor. Yes, this complacent approach to learning when you are expected to be the paragon of knowledge in the highly esteemed position as an assistant professor is unacceptable. This is your student loan dollars at work. Wow.

[@pancakes99ify]

Wow. Brice is pretty emotional 😂😂 I'm sure a joy to be around. What a mean saaaaad, nasty individual

[@brianfoster8108]

I bet that felt good, Brice. Nothing like smug self satisfaction. Judge much? But I think you missed the point. This is an introductory talk that invites users (probably beginners to intermediates) to check out forcats. I actually learned something. If I were a student...in....your....clazzzzzzzz. Sorry I nodded off there for a minute. Brice, you are clearly fairly well versed in stats and R, but rather than convey the weight of authority, your rant reeks of immaturity and churlishness. Carry on, Amelia.