Comentários e feedback de alunos de Modelos Regressivos da instituição Universidade Johns Hopkins

Requires much more than a month to digest the material and complete the assignments. A default/initial one-week offering is too tight unless you are only taking the course (not working). I know one can complete the course in more than a single round and I did that but I still don't think the expectations should be set for a single month.

Instruction (video content) can be much better, at least compared to a lot of other courses on Coursera.

Expects a level of statistical knowledge already.

Better than Stat Inference, and gave some reasonable intuition, but could be improved I think by focussing on more understanding and less maths and formulas. Some of it did seem to be - here' s a formula, plug the numbers in to get the quiz question right, whereas in reality (in the world of work) that question is completely unrealistic - you have raw data and you need to do the regression and understand what it means.

To me, the approach was too much from the theory of statistics and its mathematical foundations; I would have appreciated a more applied approach for this course in the specialization. So starting from examples, questions anout data and then working towards theory instead of the other way around.

The course is good. Unlike the previous courses I took in this track, there was almost no interaction in the forums and also no-one to give feedback. I wonder if there were any TA's present in this run.

I enjoyed the earlier courses more. I did not like the way the materials were provided. There were a lot of very complex ideas were presented, in a very concise and brief manner. Also, there should be more exercises to practice. May be its me, but, I guess, I might need more time to fully comprehend the materials.

the lecture notes is a bit confusing

Lots of room for improvement on this course, the teacher really seems like he cares but he is a really bad teacher nonetheless. The course material is incomplete and not properly structured. Basically read the book if you want to learn something, otherwise the videos don't really help.

Also, the course project is not worth it because you get no real feedback to compare your project to the ideal or at least expected answer. I would not recommend this course.

Not impressed. Dr. Caffo tried to use non-calculus, non-linear-algrebra ways to explain complex concepts and derivations. IMO, he should not have done that. It only made things more confusing. Also the final project is so unsatisfactory in that we were to analyze the data with 32 obs but 11 variables! How robust could it be? Was expecting something much more challenging than that.

I think this course needs more emphasis on practical applications and less mathematical background.

Given the importance of this subject, this course should have been split in two or more or have a longer duration to properly address subjects as GLM or model selection techniques.

The instructor is at time incomprehensible. It would be helpful to speak more slowly and pause more often. Otherwise he sounds like repeating something that he's so well memorized after many years of teaching.

This course is better than Statistical Inference, and I think it is as useful. Non credit excersise are still very good at helping with understanding in practice what is going on.

This course is great, instructor is good, however, the material of this course is not well organized, even the swirl practice is not put in the correct week, not in the same pace as the lectures. The quiz and project are far much easier than lecture content.

The course was of average quality. It could have been better. Brian's slides in the video don't correspond 1-1 with the slides made available. The coverage and explanation of the material could have been better. The instructor's presentation could be more engaging (fewer 'ums' while talking). It was not immediately clear how to answer some questions on the Week 4 quiz, and also the course project, even after reviewing the material multiple times. One example: Brian says that the ANOVA test can only be used to compare models, when the model being compared has normally distributed residuals (using the Shapiro test). No advice is given about what to do if they are not normally distributed, which is what happened in the project.

As for the Statistical Inference course, this course is amazing but is presented in a more complex way than it should be. Once again the concepts are simple and the math not so hard, yet I had to do a lot of research outside the course to be able to understand these simple concepts and derive the not so hard mathematics.

Brian Caffo is clearly brilliant and, I would say, seem to be a good lad too, but something is missing. Too often the details are thrown at us without being properly framed in the context or without having the proper concept being introduced progressively.

I have a theory about teaching since I was 15, and so far it has proven to be true. Imagine that learning is about climbing a mountain in which tall steps have been carved. Each step is taller than the student. The teacher is somewhere higher than the students (not necessarily at the top, if there is such a thing).

The job of the teacher is to throw boxes (concepts) and balls (details) of different size, shape and colors. The job of the student is to catch these boxes and balls and to put the right balls in the right boxes in order to make a staircase out of it to climb (at least) one of the giant stair up.

A good teacher makes sure to throw the concepts first than the details and to clearly specify which balls go into which box, as well as which boxes go inside/over which other boxes.

But most teacher simply throw the balls and boxes in an not so well structured manner, so the poor students try to catch as many as he can, but also miss a lot of them. His hands can hold a limited amount of balls. If he doesn't have the right box to put them, he would either miss the next balls, or put the one he hold in his hand in the wrong box.

Bottom line, the best teachers are those who focus on the concepts (and context) and make sure that the concepts are well understood before introducing details to stuck in these concepts. From my experience our brain (or at least mine) better learn this way. It is as if our brain need first to establish a category-pattern (the concept/context) to which it will associate detail-patterns. But without a proper category-pattern, our brain is having a hard time to properly remember the detail-patterns or miss-associate them to the wrong category-pattern (which create even more confusion).

Hope it was helpful somehow...

I find it very tough to understand everything. Buying the course book helps to overcome this. You have to dedicate a lot of time.

There's just something about the course content that is difficult to attain. It's presented at way too high of a level without enough tangible examples of getting down into the weeds of how to actually perform and interpret the models and functions.

A good (although slightly frustrating) course, attempted once but had to come back after studying the material in class, quite a heavy course if you've not been taught regression before

good quick overview, could have more actual R examples in lectures

Many things are not clear enough in multivariable regression part.

Useful class, but the content often simple in nature was explained in a confusing/complicated way. But the material is important and there is purchase for taking the class

Course has more theoretical concept than application.. It has to be more application based

Pros: The instructors of this course are absolutely knowledgable on the content here. The content itself is challenging and applicable to real-world data science challenges. Using R makes this a good course for today's (2019) current programming world as many professional statisticians will use this language day-to-day.

Cons: The content feels mismanaged. Sometimes the Lectures don't prep you for the practice assignments, and sometimes neither of those prep you for the quizzes particularly well. I had also hoped for some more engaging video content from a course this expensive. Having a professor in his office hastily work through material while there are police sirens outside isn't exactly pro-level instruction (It is in Baltimore, so I get it).

Overall, it's worth it if you've got the time to power through relatively dull lectures. The R based practice assignments are wonderful and the final project incorporates things together nicely.

One of the most difficult courses in the whole programme. From my point of view it is very important, but not so well explained. I had to go through other training sessions in order to understand the concept based on numerous practical examples and then return to Coursera to finish it up.