ETC2420: Statistical Thinking

Computer Lab Exercises 20% - each week you submit a computer lab task in groups of 3-4 by Tuesday 4pm that is carried out in R. These tasks varied in difficulty, to being very easy and straightforward to quite complicated, especially in the Bayesian inference section. However, having some coding experience previously, I found these reasonable to work through, especially because the R interface is setup very well. The best 10 out of 11 of your computer labs contribute to your final mark. Overall, most groups did reasonably well in this, particularly due to the vast amount of information available online to help complete the labs.

In-workshop exercises 10% - In the workshops, there was always a task to do in the middle of learning content. Sometimes it was just a MARS poll, so participation was all that was needed to get the mark for the week. However, there were also quizzes and Rmarkdown files to download and complete tasks in, that required a bit more effort and time to get the marks. These all had to be completed within a set time limit, so it was always difficult to get the tasks done in time. Also, one of the R files had issues that were not noticed before the task was attempted, leading myself and one of the tutors assisting in the task to have to solve the code in lecture, which was quite annoying. That being said, as long as you've acquired some basic knowledge of R by week 2/3, then these tasks aren't too hard and it is relatively easy to get 10/10 for this part. Similar to the labs, the best 10/12 workshop scores count towards your final grade.

In-tutorial exercises 10% - Similar to the workshops, there are also quizzes/Rmarkdown files to complete in the tutorials themselves. These can be quite repetitive, but they are usually out to trick you so they are not too difficult to do. You work through these in groups, usually with the people that you are sitting with on your table, but class interaction is encouraged, so you can generally check with others around the room to make sure you are solving these tasks correctly. Like the workshop quizzes, they didn't seem to be too daunting, but they still have some tricks here and there, so it is crucial that you check through your answers before you submit them onto Moodle, as there are often tiny tricks or mistakes in everything. Like the workshops, the best 10/12 marks in your tutorials contribute to your final grade.

Exam - 60%

To be honest, coming out of this subject I don't feel like I learnt that much new information. The course starts off with basic randomisation & simulation tasks, which are very similar to those done in VCE Methods/ some parts of ETC1000. Thus, if you have a good knowledge of these, then this part isn't that difficult, as it even goes to as basic as tossing a coin and working out what is a random throw of a coin and what isn't. The next part of the course is regression modelling, which covers model selection and interpretation, much of which is basically repeating what is covered in ETC2410, particularly with AIC/BIC/adjusted R² model selection. There are some new parts to the course, but these are relatively straightforward and easy to get your head around, as like the parts above, they are just model selection areas, so they shouldn't be too difficult to do, even if you've only done ETC1000 before doing this unit.

After this, the unit delves into Bayesian Inference, which to be honest should've had much more time allocated to it than just the 3 weeks that it was taught in. You start off with Bayes Rule and basic application of that, but then this evolves into prior/posterior models, Bayes estimators and loss functions. All of these are taught relatively quickly and not that concisely, so reading through the lecture slides and asking questions in the tutorials about these parts is paramount to understanding a full knowledge of these parts of the course. I had Lachlan and Frank as my tutors, and they were both very well learned with these parts of the course, so I'd recommend for sure to have them as your tutors if you're doing the course next year. From what I know though, the course is undergoing a bit of a revamp due to this being the first semester that Catherine has ever taught the subject.

Also, the unit heavily uses R in semester, although there is no R coding in the final exam. For most of the people studying the course, this is their first time using R (or even coding in general) , so it does seem to be quite a learning curve to understanding how to properly code in R. However, I had done some coding in the past using Java, so I found the learning curve to be not as steep, even though the R coding was basically taught via a "Google it" approach if you didn't know how to code something . Despite this

- The exam consisted of 2 sections. The first section was 10 multiple choice questions, which were very similar to the sample exam questions, and were mainly designed to trick you as much as possible. The second section consisted of 4 questions, which ETC2420 students only had to choose 3, whilst if you are doing the masters unit ETC5242, then all questions needed to be completed. I found the exam very straightforward, as it was very similar to the sample exam, such that I felt like I was repeating my sample exam answers over and over again. Since only 3 out of 4 questions in section B had to be completed, I decided to skip the last question on Bayesian inference, despite having invested so much effort into studying for that section in-semester, as I found working on the CDF/PDF relationship much easier due to doing this unit at the same time as ETC2520. Due to this, I guess that the exam was designed in a poor fashion, as it made it very easy to score high marks due to the relative simplicity of the questions being asked. This may have resulted in the marking being done in a very harsh manner though, which led to my result being a bit lower than I expected. I expect that the exam may change in the future though, as from what I know most students came out of the exam feeling that it was way too simple.
Overall, this is not a difficult unit, so it is quite easy to do decently. The only reason I deducted marks from my rating was due to the way the course was taught, so if you are planning to take this next year (as it is compulsory for a data analytics/ actuarial science student for exemptions), then I'd recommend trying to practice some R coding before you start the unit, as once you realise that you have a 3-4hr lab to complete each week, then it is quite difficult to code unless you have a general understanding of R functions as a whole, even though they do teach you from scratch how to code. Also, from what I know, as said before the course is undergoing a revamp, such that some topics will be removed and a greater focus on Bayesian inference will be added. This will hopefully mean that it should be much more straightforward in the future to understand the coursework. That being said, this was my favourite unit of the semester, due to the relative ease at understanding the content, and my level of engagement in the R coding tasks.

Catherine Forbes - Also chief examiner and unit coordinator. This semester was her first time taking the subject, so she did seem to struggle a little bit with teaching the content. That being said, I found the workshops she taught reasonably engaging due to all of the interactive activities that took place in them

One sample exam provided with solutions, but past exams are available online

4 out of 5

Yes, with screen capture

No textbook required as all resources are provided in Moodle including necessary textbooks for free.

- The tutorials are done in the large rooms of the first floor of the LTB. They heavily focus on group interaction, which is great as I felt that I learnt a lot more in them than the workshops due to their collaborative setup. After doing the tutorial exercises that are explained above, the rest of the tutes are mainly dedicated to working on the computer lab exercises for the following week (if there's any time left in the tutes to do them). Therefore, they helped out a lot in understanding what the computer lab exercises were asking, as well as asking your tutors for any questions that you had about the content in general. Due to the tutorials consisting of about 40 students per tutorial (leading to there being only 5 tutorial options as a whole), then there are 2 tutors who are able to assist you if you have issues with coding the lab exercises for the following week or completing the in-tutorial exercises. I found these very engaging, as you had more personal interaction with the tutors to understand what was going on as a whole.

1x 1.5hr tutorial per week1 x 2hr workshop per week

Semester 2, 2018

87 HD