Changing my college major to statistics -- Should I?

[Felipe Lincoln]

I am current in the second year of college, doing physics but from the mid of the year I got hooked to data science and I decided I want to follow this career. What is the best thing I can do now? Finish my BSc and go for a MSc in statistics? Change my major to statistics? As I still like physics should I find a area that I can apply machine learning and finish BSc and go for MSc in physics?
I'm a lot confused now. The only thing I know is that I love dealing with data, graphics and I like science: physics, biology even medicine.
Can you give me some advice on how I can "find myself" ?

 

[ZeGato]

Finishing your BSc and going for a MSc in statistics, or data science, sounds like a good idea; many people working in data science today come from a Physics or Maths BSc without even having any background in Statistics. Changing your major now would mean that you'd take a few more years in university, and with what purpose? A masters will teach you enough of statistics to apply in your job, and you'll also learn a lot of data science in your workplace. University isn't the end of it, just the beginning.

What you can do, if your university allows it, is to take a Minor in Statistics, while in Physics? That way you don't have to lose any precious time, while learning statistics along the way.

 

[Stephen Tashi]

but from the mid of the year I got hooked to data science

How did this happen? One course from a dynamic teacher? Independent study? A class project?

 

[Felipe Lincoln]

How did this happen? One course from a dynamic teacher? Independent study? A class project?

I began a research in machine learning and then started competing at Kaggle

 

[Felipe Lincoln]

What you can do, if your university allows it, is to take a Minor in Statistics, while in Physics? That way you don't have to lose any precious time, while learning statistics along the way.

Yes I can do it!

 

[Dale]

Changing your major now would mean that you'd take a few more years in university,

Not necessarily. My wife had something like 7 majors and still graduated on time. Changing majors may not delay graduation, but it will reduce the number of electives.

 

[Felipe Lincoln]

Not necessarily. My wife had something like 7 majors and still graduated on time. Changing majors may not delay graduation, but it will reduce the number of electives.

wow that is impressive!
But what do you mean by "graduated on time" ?

 

[Dale]

wow that is impressive!
But what do you mean by "graduated on time" ?

She finished in four years

 

[FactChecker]

If you love statistics and data analysis, switch and don't look back. You are in your second year and it sounds like you already have some good experience and classes in statistics. You should not loose much time by switching.

 

[StatGuy2000]

To the OP:

If you still like physics and wish to continue your studies in it, I don't necessarily see the need for you to switch programs. A physics degree provides you with the quantitative background that graduate programs in statistics typically look for. That being said, if you do also enjoy statistics and data science as a field, I would certainly recommend taking more courses in that field.

My suggestion would be one of the following:

1. Continue your current BS program in physics, with a minor in statistics, and then pursue a MSc (or perhaps even further graduate studies) in statistics. Make sure during your studies to add a few computer science courses as well if you haven't done so already (since computing is an important component of both statistics/data science and physics). If you find that you are able to do so, maybe see if you can double-major in physics and statistics, although I would be careful about not overextending yourself.

2. If you absolutely love statistics and data science, and don't feel the need to go any further in physics, switch to the statistics BS program and then continue graduate studies in statistics. Only do this if you know you won't regret not taking further physics courses.

 

[Felipe Lincoln]

Much appreciated for your post @StatGuy2000 and @FactChecker !
Right now I feel a would be more confortable working on a problem in science through data collected from experiment and previous knowledge about the problem. But I do not know if I would like to work on a pure statistics problem in case I went to a MSc in statistics and have to work on a thesis.

 

[Felipe Lincoln]

I also think of the possibility on working out of academia with data science and have physics as a hobby.
I love data science because we can work in almost every field we want, like oceanography, geology, archaeology and so on. Learning things about lots of different area inspires me.

 

[FactChecker]

IMHO, every interesting problem has some statistics (due to random and/or unknown factors), some optimization, and some feedback/control.

 

[Locrian]

IMHO, every interesting problem has some statistics (due to random and/or unknown factors), some optimization, and some feedback/control.

Yes, I like this a lot. I don't think it's so applicable to the OP, but I've started pointing people to Operations Research studies as well, as I think there's some interesting math, computer science and problem solving in that field. It's also almost perpetually in demand, and due to low awareness doesn't seem a common area of study.

 

[Felipe Lincoln]

I never heard of Operations Research before, but it seems interesting at first sight. I will do some more search about it.
Thank you !

 

[Felipe Lincoln]

I don't get it why there is more computer scientists in data science than statistician. What is the main reason?
My university allow us to get even two minors, considering this scenario what about staying in physics and chosing cs and statistics minors instead of going to statistics major?

 

[FactChecker]

I don't get it why there is more computer scientists in data science than statistician. What is the main reason?

There is usually a lot of work involved in collecting data and getting it into a usable form. A computer person is best at that. Once that is done, it requires statistical knowledge to analyze the data. If it is possible to design experiments to get the data, that is also a statistical problem.

My university allow us to get even two minors, considering this scenario what about staying in physics and chosing cs and statistics minors instead of going to statistics major?

That is up to you and it depends on what subject you like the most and what career you are interested in.

 

[Locrian]

And it's possible to get a masters in statistics having spent very little time with the tools used in data science. Skip a few classes and you'll find yourself graduating without having ever used xgboost, or able to walk someone through the difference between ridge and lasso. We know this because we've seen this in interviews!

Really, that's true of computer science, too. Analytics has, in my opinion, largely split from both statistics and CS. This doesn't mean people with those degrees don't understand or have knowledge of modern analytics, but it's not a requirement, either. It's more of a foundation, and either one can provide that.

I will say that in some ways CS provides a bit better a background. Many tools used in predictive analytics are very distant from traditional statistics, and the OR side of prescriptive analytics has nothing to do with stats. Yet no matter what you're doing, you still have the problem of how to leverage the tools against large, sometimes complicated data sources. Obviously there's some critical skills required to deploy the right models, and feature engineering is still very much an art. But an awful lot of the actual work comes down to building a robust process, and that's closer to software engineering than it is statistics.

 

[FactChecker]

Yes, I like this a lot. I don't think it's so applicable to the OP, but I've started pointing people to Operations Research studies as well, as I think there's some interesting math, computer science and problem solving in that field. It's also almost perpetually in demand, and due to low awareness doesn't seem a common area of study.

I think that the only major thing missing from my OR classes (long ago) was the subject of feedback and control. That is a very significant component of many problems.

 

[Locrian]

and the OR side of prescriptive analytics has nothing to do with stats.

That's a silly exaggeration of mine by the way, but hopefully the overall point got through.

 

[StatGuy2000]

I think that the only major thing missing from my OR classes (long ago) was the subject of feedback and control. That is a very significant component of many problems.

@FactChecker, as an aside, the subject of feedback and control is considered the preserve of control theory. I'm curious if control theory is considered part of the curriculum within operations research (either at the undergraduate or the graduate level).

 

[FactChecker]

@FactChecker, as an aside, the subject of feedback and control is considered the preserve of control theory.

That is my experience also.

I'm curious if control theory is considered part of the curriculum within operations research (either at the undergraduate or the graduate level).

Not that I know of, but I think that so many optimization problems are also control problems. In OR, there were endless simulations where there was significant feedback, but the subject was not studied where I was (40 years ago).

 

[StatGuy2000]

Not that I know of, but I think that so many optimization problems are also control problems. In OR, there were endless simulations where there was significant feedback, but the subject was not studied where I was (40 years ago).

I find that interesting, because in the industrial engineering/operations research departments in a number of prominent schools (including both my alma mater and Berkeley, for example), control theory is an important research area within those departments.

See for example the link to the Berkeley OR department:

https://ieor.berkeley.edu/research/areas-of-research

Perhaps this may have been the situation of OR departments in the past?

 

[FactChecker]

I find that interesting, because in the industrial engineering/operations research departments in a number of prominent schools (including both my alma mater and Berkeley, for example), control theory is an important research area within those departments.

I'm glad to hear that. It's smart.

See for example the link to the Berkeley OR department:

https://ieor.berkeley.edu/research/areas-of-research

Perhaps this may have been the situation of OR departments in the past?

That is possible, or I may have just been unaware of it. The core classes in the department did not include it.