STAT 3220

Whew. Krista is so difficult. While I will more than likely get an A- only because her grade thresholds are ridiculous, she is so unaccommodating. She expects us to read the textbook, watch pre-lecture, go to class every day and complete classwork (graded for accuracy) have a lab and homework (so long) for every unit, and 2 projects, it is just way too much. Not to mention, she isn't very approachable, does not allow extensions, terrible communication by email, and isn't a good lecturer, and there are barely any drops. If this class isn't a requirement, take an alternative she is such a piece of work and it is not worth your stress. It is unfortunate because I took this class during online learning, so I can only imagine what it is like in person. The only good thing really is that there are no exams, and you work in groups for class work and the projects.

Honestly Varanyak gets a lot of unwarranted flack for this class. This class is not difficult as long as you skim the textbook and take notes on the weekly lectures (only like 30 minutes each). Yes, we don't go in-depth to the mathematical concepts - that is outside the scope of this class. You will probably get more out of this if you've already taken STAT 3120. This class advertises basic multivariate regression skills. You will learn to do that.

The grade threshold for this class is a 95, but with how easily most assignments are graded it is not a heavy lift.
The problem sets are easy and are direct applications from the classwork and lectures.

I will concede that SAS sucks, but you can literally just copy and paste code from the provided examples and refit it to your context with extremely limited/no programming knowledge.

Krista throws a lot of information and assignments at you- prerecorded lecture videos, daily class meetings, classwork assignments, lab assignments, monthly homeworks, and project deadlines. However, most of this work is really manageable and can actually help enhance your understanding of linear regression if you try on them. The course is an easy A if you use the examples that Krista provides and apply them to your homeworks and labs while changing the numbers. The TA's also really don't know what's going on but they grade easily. The final project can be fun and accomplished pretty quickly if you pick your group wisely. The material in this class is definitely relevant to the real world, let alone a pre-requisite for almost every concentration of the stat major, so I definitely recommend taking it.

Learned a lot of useful methods for analyzing data and building multiple linear regression or logistical regression models. The grade is mostly based on group work so it is important to pick a good one. There are lots of little deadlines and you don't want to forget about them as their is no late policy.

I think the past reviews of Krista were rather harsh. I think she's changed the format a bit since previous years. There really wasn't a lot of homework most weeks. She uploads lecture videos that we're supposed to watch before class, and they're helpful for the in-class group activities. The lecture video slides are thorough enough that you don't even need the textbook. She spends some time at the top of every class going over the same material as the lecture video which I rarely paid attention to tbh, I'd just get started on the in-class assignment. The in-class assignments weren't difficult, and she basically provides all the SAS code you need so it's fairly straightforward. The "homeworks" were completely optional. Instead, we had weekly quizzes on Gradescope that weren't incredibly difficult either if you looked at the lecture slides. Each unit there would be a "lab" that you'd work on with a partner that was essentially a longer version of the in-class assignments. The most stressful part of the course were the two group projects. Because the day-to-day class activities were so simple, it felt harder to actually apply the concepts we were supposed to learn to the project. My advice is to not procrastinate, get a good group, and go to office hours to ask questions. Krista and her TAs were generally pretty good about answering questions. One other gripe I had with the course was that an A was a 95 which is needlessly high imo

Krista has DEFINITELY listened to reviews and has changed this class for the better. A lot of the complaints listed in reviews from previous semesters have been fixed. Homeworks are optional, there is no busywork, and the workload is extremely minimal. On average, I spent less than 1 hour working on this class outside of lectures. Instead of homework, we have 10 weekly quizzes that are a total of 15% of your grade. She drops the lowest 2, and the last 2 are automatic 100s (for completion). I thought the material was interesting and Krista was super nice and approachable. The bulk of your grade in this class will be from the project (20% for part 1 and 25% for part 2). The rubrics are straightforward, and I found it very easy to be successful in this class.

I really enjoyed this class! Professor Varanyak used a "flipped classroom" where she posted video lectures we had to watch before the class, then reviewed the material in the first thirty minutes of class, and we spent the rest of the time working on group work, short daily classworks based on the material we had learned. SAS was really tedious and irritating to learn, but she gave us more or less all the code we needed. We had "labs" for each unit, basically a larger classwork with material from the whole unit, which was in partners, weekly open note quizzes, and a large final project split up into two parts. To anyone considering taking this class, I say go for it, its not that bad, and focus most heavily on the quizzes, labs, and project - don't slack off. #tCFfall2021

(+): easy A, organized materials, approachable professor/TAs
(-): graded in-class works (=participation is required to get grade), all team based works

Take it if you want easy A
Do not take it if you don't like to go to class or prefer individual work

Don't get me wrong, regression analysis is a crucial part of Statistics, and this course is very foundational, albeit barebones. However, it's really not as big of a deal as the Professor Varanyak makes it seem. She also doesn't really try to make this class interesting.

There's a really big ego coming off from her. She seems to think that this is the biggest class we're taking this semester, and our absolute priority. Assignment extensions? For what, other classes? Why aren't you prioritizing this one? What, you don't know SAS? Here's a 20-hour tutorial for the programming language that you'll only use for this class, because R is better in virtually every way, except maybe one or two cases. What do you mean you don't have time for a 20-hour tutorial? You only have 3 lecture hours a week.
She uses the "flipped" classroom style, giving these worthless presentations that don't advance your knowledge of the material at all. People caught onto this pretty early, and she openly complained in class that no one watches her lectures (I wonder why). She is extremely unaccommodating, and is useless in OH, basically agreeing with whatever you say, even if it's very much not right. On that note, she also complains that no one goes to her office hours (I wonder why).
There are a few diamonds in the rough, however. She posts a lot of extra practice problems (with solutions), and assignments/labs in this class often are very similar, so it's not like she completely leaves you in the dark. There is a lot of structure to this course, and she's pretty good with following the schedule she establishes at the beginning of the semester.

Let's talk grading:

Assignments: 5%. You can copy/paste SAS code, with a few numbers changed around, and you're practically guaranteed a 99%. She drones on through SAS code for 45 minutes, do the assignments during then, and you can leave early 98% of the time (she won't give you enough time to finish the assignment in class if you actually pay attention to her lectures).

Quizzes: 20%. You're guaranteed to miss at least one question on these, because these questions are hyper-specific, but since Regression is your only class of course, you should be fine, since you've obviously memorized every word she says in her VERY pointless recorded lectures (she might say the answer in these, like once. Maybe). One of these grades is to got o events related to the Statistics department

Labs: 30%. Basically a harder assignment that takes the entire class time (she won't lecture on these days). They're pretty doable if you look at the assignments.

Project: 45%. She should try out for the cheer team, the way she stretches to mark off random points on this.

Overall, get ready for a professor who acts like she's spearheading the revolution in statistical learning. This class could've been so much easier if she didn't act like this was our only class.

#tCFspring2022

Understanding the concepts of regression is absolutely crucial to statistics, and if taught well, can very effectively be used as a "bridge" to connect introductory statistics courses to higher-level "machine learning" ideas (e.g. the sigmoid function used in logistic regression is also a common neural network activation function). The idea of a hands-on regression course with plenty of practice with linear, multiple linear, and logistic regression is very sound, and when I was actually doing the coursework for this class, I found it to be extremely relevant and useful to future statistics courses I may take. The in-depth exploration of the regression assumptions in STAT 3220 stood out as one of the course highlights — understanding concepts like homo- vs heteroscedasticity & interpreting Q-Q plots is VERY handy when actually applying regression techniques in the field. I am very glad that UVA offers this class & think the overall material is incredibly valuable.

That being said... the actual experience of this class was very frustrating, for a few reasons:

1. The entire course is taught using SAS, a dinosaur of a programming language that is almost totally obsolete outside of the field of biostatistics and is a total pain to learn. It's so convoluted that the instructor required us to complete a 20-hour course on the basics of the programming language and beyond the basic proc... and data... syntax I'd be hard-pressed to pass any sort of programming exam on the language. It's inflexible, doesn't really support functional or OO programming, and has horrible documentation. I legitimately cannot imagine the justification for teaching a statistics course in SAS instead of R or Python in the year 2022, especially because SAS is not open source and charges an excessive amount of $$ for access to an online web editor (that's right, you can't even program in SAS on your actual computer, you have to use an unresponsive and buggy code editor in the browser).

2. The feedback on assignments ranges from "pretty useless" to "totally nonexistent." Points are taken off arbitrarily on labs and even MORE arbitrarily on the final project, which — despite having multiple full class days devoted to it — was never really clearly defined. Some cautions on the labs: while during the lectures and homeworks, the professor emphasizes the arbitrariness of the regression model building process and the lack of black-and-white answers in variable selection (the "garden of forking paths" is a familiar concept to students with a statistics background), do NOT deviate from whatever the graders have on their answer key — no matter how far you go to try and explain your reasoning, it won't be far enough! If you're worried you don't understand the content covered in video lectures, some copy-pasting of the HW code + changing of numbers is good enough to earn As on the labs.