Students of STAT 101!
When I designed STAT 101, my hope was to make the material accessible to everyone, even students who don't like math. Statistics, to me, is a highly intuitive field, a field full of magic and surprises. I aspired to share these insights with everyone, and have students experience them by themselves, by working on interesting problems. In making this class, I also experimented with a mix of materials. Plenty material is optional, such as mathematical proofs that students can elect to construct. We have been carefully studying engagement numbers and student feedback for these experiments, learning a lot about how students learn in this new online medium.
We have also received a lot of direct feedback from our students throughout the course in our forums and in discussions which we really appreciate. And I recently came across this this article by another professor, which I read very carefully. While luckily not all of our students share this view, this article points out a number of shortcomings that warrant improvements that a few students have also raised and we have noted for iteration. Some are the result of my attempt to experiment with open questions (challenge students before I provide an answer), my dedication to get rid of overly formal definitions, and my desire to place optional challenges into an otherwise basic course. But I agree with the author of this article -- and our many supporters who have voiced similar things -- that the resulting course can be improved in more than one way.
Our online classes are being revised frequently. We use the data and feedback in this medium to adapt and further optimize. In the next weeks, we will majorly update the content of this class, making it more coherent, fixing errors, adding missing content. I believe at Udacity we owe all our students our hardest and best work in making amazing classes. We are very grateful for any feedback that we receive. These are the early days of online education, and sometimes our experimentation gets in the way of a coherent class.
For clarification, the "final exam" in this class is for students to prove to themselves that they got the material right - not as a test that independently certifies a student's performance. We have launched an proctored exam for CS101 and in time will do the same for this class.
PS: Please fire off additional suggestions for improvement in this thread.
asked 11 Sep '12, 14:00
In the square feet/expected price video, the answers do not use commas - 144000 - but the closed captions at the bottom of the screen do - so when I answered 144,000, it was graded as wrong
Please consider making the screen and closed captions match (both no commas), or don't allow the answer box to accept commas, or add a small note that in stats you don't use commas, or a small note that "no commas in answers" before someone does it wrong.
Granted, none of the written numbers in the video have commas, but I don't think that means every student every time is going to notice that. Plus, everywhere else in life, people commas in large numbers.
Is this major? No. However, a little fix will save maybe the 10 to 15% of people who might add a comm by habit.
Also, somebody mentioned that the same questions get asked over and over in the discussions boards and everybody should read all the posts before asking a question. Is there a way the discussion board can have "drop down" box of that shows the same string of words or key words as you are typing?
i.e., as I type "How do you figure the mean"
I get a drop down box with
How do you figure the mean for Question 3
of previous questions ? This may prevent the same question being asked 9 times.
Obviously, there ways to ask the question, so may key words may also help.
What's the mean for Question 3?
Perhaps the keywords "mean" and "Question 3" would show the question has been asked , or asked and answered.
answered 05 Apr, 00:59
I think the article misses the point. The author objects to MOOCs for not being a traditional course. But MOOCs are something new, and they fill a real need.
I did Professor Thrun's course as a refresher. I liked it a lot, and think it would also make a great introduction. I think the breezy "let's get on with it" style is really useful for demystifying the subject and making it accessible.
If you are still considering updates, I would really like more about how you apply statistics in e.g. your robotics work. There was a brief mention early on where you compared the imperfect test for an illness with the imperfect sensor in a robot. That was interesting! Could you have any more examples like that? Maybe they could get joined up into a mini-project. Udacity courses are great at showing how mysterious real-world problems (saving Apollo 14 astronauts in CS222, competing with Google in CS101) can be solved with a little training and a few weeks work.
answered 05 Jan, 17:53
As a software developer, I found ST101 very rewarding. The approach taken by Sebastian and team, to bring statistical thinking tools to life for a broad range of students, is the one that is going to do the most good over time. Critics who think that a more formal, dry approach with a more rigorous Final Exam would be better are missing the big picture in math education: The top-level goal is to increase fondness for the topic (or to allay anxiety surrounding it). The most important metric is future affinity toward the topic: Do ST101 students emerge more comfortable with suggesting and using statistical techniques in problem solving (than before the learning experience)? I am definitely more turned on to statistics having just completed the course.
I learned much more about Python as a result of doing the Programming Exercises. Having the programming be optional is a good way to serve students who are ready to explore computational statistics, while respecting those who are not.
One suggestion in regards to the Bayesian Inference topic would be to choose a simple example from machine learning. Young people deserve to know that, during their lifetimes, learning machines will be one of the coolest applications of statistics! One of the greatest lasting impacts of this course will be to expand the learner's sense of career options.
The Monty Hall problem was terrific. When I wrote code for how the MH Show operated, I slipped into the role of a game show designer. Wow.
I would definitely take a ST202 course if offered. A great programming lab for that course would be designing a Gaussian Data Generator, and then designing a test strategy to evaluate it. Also, if you can offer a Python 2D graphics canvas to students, it will open up many interesting possibilities for graphic simulations. A 2D random walk would be one of them.
The critique appears in the blog AngryMath. Hence, readers should recognize that the author is venting. Notice that the author's primary blog is about the game dungeons and dragons ("D&D Hotspot"). Personally, I applaud Professor Thrun's intuitive approach, and I like the way the course attempts to build understanding of concepts without being pedantic. For instance, I loved the use of the binomial probability distribution and small sample sizes while teaching the concept of hypothesis testing.
answered 03 Jan, 19:40
hello professor sebastian,
I am enjoying your course, but I agree with the mentioned article: it can be much improved. I believe you can look for inspiration in two books:
I actually read most of the second book, its surely useful. I've never read the first one, but it seems appropriate, since its introductory and about python and statistics.
I'd also like to suggest that you take great attention to expository details. For instance, having the questions of quizes written on the video/comments would be very nice. I took CS101 and it's a good example to learn from.
I'll surely check out the course redesigned. Online education is the future, so I'm really excited about participating in its blossoming. I wish you guys the best.
answered 28 Dec '12, 17:40
First, an instructor at any level does not cause a student to learn. In fact, the more skilled an educator is at inducing learning, the more risk that the student will become dependent. There is no magic formula--an infinite variety of teachers could possibly inspire students to want to learn. The desire in the student is the cause. Even an inept (but perhaps likable) instructor could inspire students to want to learn, if only out of pity or sympathy. An obnoxious or arrogant instructor could effectively inspire students to want to learn, perhaps to prove they are better than the teacher. Any student with their own reasons to want to learn the material can find a way to overcome whatever limitations of the teacher.
Maybe a better approach is to seek to make the material engaging and entertaining, like a game or song or movie. There is some craft to that, but creativity is the key to building a "hit" from an original concept. The unique qualities and characteristics of a particular instructor (including assistants) is actually an asset in that regard.
I don't profess (pun!) to be qualified to write a hit online course, but I did have an idea that might illustrate my point. You begin the added Unit 11A with a question about a nerf gun projectile. Ultimately, the objective is to give the student a gratifying experience of a realization of the conceptual basis for probability density. I hope I understand well enough, but bear with me.
Suppose you begin with the question about the spinner, by asking "What is the probability that the spinner will land within one particular half of the circle?" Then within 30 degrees, 1 degree, a fraction of a degree. The student can be led by the expectation that the range of possible results will continue to decrease in size to the conclusion that the probability of a specific angle is immeasurably small--effectively 0. This gives the student an experience of success, however contrived, but he can then be provoked further by the question "Why is that?" or "What does that mean?" The ensuing lecture content can explain that in each case the probability applies to a range of values, and surprise the student with a context-shift in the form of re-framing the probability of the result falling within the range between a specific value and itself: 0.
This kind of leading the observer (audience) down a path to a conclusion followed by a twist or surprise is a common feature of all forms of entertainment (comedy is a great example) that could be used as a launchpad for a creative process of developing course content.
I have to say that I think that point 7, "Bipolar Difficulty," is the one I have found most frustrating. Questions range from so simple that it isn't worth the time to type the answer to I don't know what is being asked.
The criticism on the final exam I found the poorest. Sure it is easy to cheat on Udacity. The only person harmed is the cheater.
answered 07 Dec '12, 21:50
Although I'm no expert in statistics, I disagree with Dan and I enjoyed the course. I've learned a lot and have found the problem sets thought provoking. I'm very grateful for this class. The bugs that he has pointed out are there, but an online class is not different from any software or other product, you will always need to iron out some bugs. I have some suggestions though:
Here is my view.
I didn't enroll in this course to learn something new (math is my subject, so basic stuff are known), but to see how a very bright scientist would explain very simple, but extremely important, concepts
I have watched a lot of people trying to teach these stuff, but not one as good as Sebastian does it here.
The whole design of the course is SUPERB. He really drives through the important concepts one has to master , to understand what is going on WITHOUT over simplyfing. Its not as easy as might one think.
The only problem , from my point of view , is that after finishing and mastering the course, a student will not be able to do any basic statistical analysis on his own. Its a great introduction to the basics, but its crying for a second part to complete the basic picture
I hope there is a second part , or more additions to this course
answered 13 Oct '12, 18:17
The only real missing part of the equation nobody's made a recomendation for so far is concerning Q+A. Not every question gets answered in the forums and not every answer is all that good, for obvious reasons. In time, the "crowd sourced" portion of this education will out-value the lectures. I believe this will be true of all online learning. Google/Wickipedia may provide correct descriptions but doesn't provide answers. These forums will need to be edited, refined, built upon and ultimatly condensed into FAQ per each topic. Hopefully as soon as MOOC's find their profit generating advertisers or other revenue stream, full time editors can work to provide each student clear directions towards the supplimental materials that best suits their learning style and personal needs.
answered 09 Nov '12, 17:32