Review: John Hopkins Statistical Inference on Coursera

So we’re finally here where the rubber meets the road. The sixth course in the series and it gets serious…

I’m not too proud to admit I found this module a lot more challenging than the others. I had to do most of the optional Swirl and homework assignments in order to gain the intuition about the subject.

One of the things I noticed about this course is that it took a far larger time commitment. I think every week tops and hour of lecture content, even in week 4; which has tended to be a wrap up and case study in the past.

Take into account the additional homework videos and optional Swirl assignments I would venture there is more than double the content than some others.


Unfortunately this is the one part of the course that perhaps lets it down. For me it is very obvious that you are watching something that has evolved over time. Unfortunately the first week is probably the poorest.

This is the first time we get to see Brian Caffo and his delivery leaves a lot to be desired. He comes across as discussing the topics with peers rather than a teacher explaining to students. He is very wooden and doesn’t look or sound like he enjoys teaching. Quite often just reading and raising his voice when stressing a key word. Reminding me of my of old school teachers.

In the following weeks and even in the homework video for week one this changes, and his style becomes a lot more amenable; with his passion and interest come through. This ultimately ending up with me changing my initial opinion of him and becoming someone I wanted to learn from.

Perhaps given that Coursera is a for profit organisation it should take a leaf from the video game industry.

Here they know first impressions count, so quite often develop the first level last – as it is the part of the game most people who play will see. It will then be their best work as they have had a chance to hone their delivery over the rest of game.

As a note if you are interested in polished delivery the course Introduction to Statistics at Udacity covers many of the same concepts and is beautifully delivered in my opinion. Especially the content by Sebastian Thrun.

Course Project

I enjoyed this, and felt it was a good length. For me this assignment was really more like what I expected from the “Exploratory Data Analysis” course. We have cursory look into a data set, where we used some canned functions in R and make some basic plots. You produce a report (ideally in R Markdown) based on you findings and asked for your own conclusions. It felt very practical and actually like a real world activity a Data Scientist might undertake.

For the first time there is a minimum word count that peer reviewers are required to enter. In the past you might get one out of four reviewers writing anything useful; typically just getting “good job” responses. This is definitely a positive for the Coursera format and my reviewers all provided constructive and fair feedback, that actually gave me the impression they looked at my assignment. In the past this wasn’t always apparent!


The quizzes are the usual Coursera multiple choice format. Personally I think they’re actually too hard, in the sense they can’t really be done without the lecture notes in hand. Then it is just a case of finding the related topic and punching the numbers into a formula. Even after doing the extra homework and Swirl assignments, I couldn’t remember the formulas required off by heart. Habitually I got in the habit of just writing functions in R that I could punch the numbers into during the Quiz.

That said; they’re just multiple choice quizzes with four invariant answers…where you ultimately have 3 retries. So it is basically impossible to get a failing score if you have some degree of sense.


Makes a welcome return after a two course hiatus, for me these are immeasurably more useful than the quiz questioning. They actually have you writing code in R and solve problems. I feel these should really be part of the official marking scheme. I would prefer to see the quizzes tackle the theory questions and Swirl to assess the student’s practical ability.

For now they are just a good way to earn up to five extra credits if you have dropped some elsewhere


These are a new inclusion to the format, where you answer questions similar to what you will find in the quiz. There is also an accompanying video where Brian explains each answer in detail.

This format works really well and judging by the production quality I gather it was probably a more recent addition to the course. No doubt (because I have mentioned) the quiz questions are pretty tough!

Final Thoughts

Overall it was a good course, evening up the score to 3 good and 3 bad in total for the specialisation.

August is busy month for me, so I’m not going to be tackling the Regression Model course immediately. I think I will actually finish up the aforementioned Introduction to Statistics at Udacity. To help further cement the intuition learned here.

Review: John Hopkins Statistical Inference on Coursera